{ "source": "https://inferencerate.io", "updated_at": "2026-10-16T08:21:21.871650+00:00", "license": "CC-BY-4.0", "count": 37, "models": [ { "id": "qwen-3-8-flash-next", "name": "Qwen 3.8 Flash Next", "organization": "Alibaba Cloud", "release_date": "2026-02-20", "license": "Qwen Community License 1.0", "context_window": 268288, "max_output_tokens": 65536, "input_cost_per_m": 0.16, "output_cost_per_m": 0.47, "cached_input_cost_per_m": 0.016, "swe_bench": 81.0, "mmlu_pro": 88.4, "humaneval": 95.1, "avg_ttft_ms": 280, "tokens_per_sec": 185.0, "ivr_score": 85.1, "wikidata_id": "Q130489955", "official_url": "https://www.alibabacloud.com/help/en/model-studio/getting-started/models", "description": "High-speed Apache 2.0 sparse MoE model with 1M context window and sub-100ms TTFT latency.", "use_case_summary": "Next-generation open-weight multimodal MoE activating only 6B parameters per token for ultra-low-cost enterprise office automation, tool calling, and multilingual repository engineering.", "last_updated": "2026-10-16 08:17:46", "cache_write_cost_per_m": 0.2, "architecture": "125B/6B Sparse MoE (512 Experts, 10 Active) with 51B N-Gram Embedding & Vision Encoder", "modalities": "Text, Vision, Video", "parameter_count_b": 32.0, "gpqa_diamond": 91.7, "math_500": 95.7, "blended_cost_per_m": 0.237, "composite_intelligence": 87.9, "is_pareto_optimal": true, "is_open_weight": true, "hosted_providers": [ { "provider_id": "alibaba-cloud", "provider_name": "Alibaba Cloud DashScope", "input_cost_per_m": 0.16, "output_cost_per_m": 0.47, "avg_ttft_ms": 280, "tokens_per_sec": 185.0, "is_primary": true }, { "provider_id": "groq", "provider_name": "Groq LPU Inference", "input_cost_per_m": 0.16, "output_cost_per_m": 0.47, "avg_ttft_ms": 255, "tokens_per_sec": 296.0, "is_primary": false } ] }, { "id": "deepseek-v4-flash-vision-exp", "name": "DeepSeek V4 Flash Vision (Exp)", "organization": "DeepSeek", "release_date": "2026-02-10", "license": "MIT Open Source", "context_window": 1048576, "max_output_tokens": 128000, "input_cost_per_m": 0.22, "output_cost_per_m": 0.65, "cached_input_cost_per_m": 0.0069, "swe_bench": 77.4, "mmlu_pro": 87.3, "humaneval": 94.5, "avg_ttft_ms": 310, "tokens_per_sec": 165.0, "ivr_score": 81.6, "wikidata_id": "Q131557008", "official_url": "https://api-docs.deepseek.com/quick_start/pricing", "description": "Ultra-low-cost multimodal open-weight model featuring Multi-head Latent Attention v4 and a 1M token context window.", "use_case_summary": "Ultra-low-cost open-weight multimodal MoE built for high-throughput visual document OCR, chart interpretation, and real-time autonomous coding agents.", "last_updated": "2026-10-16 08:17:46", "cache_write_cost_per_m": 0.22, "architecture": "285B/13B Sparse MoE with 32-Layer ViT & DSpark Speculative Decoding", "modalities": "Text, Vision", "parameter_count_b": 48.0, "gpqa_diamond": 91.3, "math_500": 96.4, "blended_cost_per_m": 0.328, "composite_intelligence": 86.5, "is_pareto_optimal": false, "is_open_weight": true, "hosted_providers": [ { "provider_id": "deepseek", "provider_name": "DeepSeek Platform", "input_cost_per_m": 0.22, "output_cost_per_m": 0.65, "avg_ttft_ms": 310, "tokens_per_sec": 165.0, "is_primary": true }, { "provider_id": "cursor-anysphere", "provider_name": "Anysphere Cursor Router", "input_cost_per_m": 0.22, "output_cost_per_m": 0.65, "avg_ttft_ms": 320, "tokens_per_sec": 181.5, "is_primary": false } ] }, { "id": "gemini-3-5-flash-lite", "name": "Gemini 3.5 Flash-Lite", "organization": "Google", "release_date": "2026-04-08", "license": "Proprietary Commercial", "context_window": 1048576, "max_output_tokens": 32768, "input_cost_per_m": 0.075, "output_cost_per_m": 0.3, "cached_input_cost_per_m": 0.015, "swe_bench": 55.4, "mmlu_pro": 80.1, "humaneval": 91.0, "avg_ttft_ms": 78, "tokens_per_sec": 265.0, "ivr_score": 81.5, "wikidata_id": "Q131505676", "official_url": "https://ai.google.dev/pricing", "description": "Google's lowest-cost 1M-token model engineered for massive-scale classification, entity extraction, and sub-80ms real-time edge inference.", "use_case_summary": "High-frequency data pipelines, real-time routing, large-scale document tagging, and ultra-low-cost conversational assistants.", "last_updated": "2026-10-16 08:17:46", "cache_write_cost_per_m": 0.075, "architecture": "Ultra-Compact Distilled MoE", "modalities": "Text, Vision, Audio", "parameter_count_b": 24.0, "gpqa_diamond": 68.4, "math_500": 90.2, "blended_cost_per_m": 0.131, "composite_intelligence": 71.1, "is_pareto_optimal": true, "is_open_weight": false, "hosted_providers": [ { "provider_id": "google-vertex", "provider_name": "Google Gemini & Vertex AI", "input_cost_per_m": 0.075, "output_cost_per_m": 0.3, "avg_ttft_ms": 78, "tokens_per_sec": 265.0, "is_primary": true } ] }, { "id": "gemma-4-31b-it", "name": "Gemma 4 31B IT", "organization": "Google", "release_date": "2026-06-01", "license": "Gemma Open Commercial License", "context_window": 131072, "max_output_tokens": 16384, "input_cost_per_m": 0.12, "output_cost_per_m": 0.36, "cached_input_cost_per_m": 0.03, "swe_bench": 59.2, "mmlu_pro": 82.8, "humaneval": 91.8, "avg_ttft_ms": 115, "tokens_per_sec": 155.0, "ivr_score": 80.3, "wikidata_id": "Q131505676", "official_url": "https://ai.google.dev/gemma", "description": "Google's flagship Gemma 4 31B dense open-weight model delivering frontier-class coding and reasoning on a single enterprise GPU.", "use_case_summary": "Self-hosted VPC inference, fine-tuned domain agents, sovereign healthcare/finance workloads, and low-latency local code generation.", "last_updated": "2026-10-16 08:17:46", "cache_write_cost_per_m": 0.12, "architecture": "Dense Open-Weight Transformer", "modalities": "Text, Vision, Code", "parameter_count_b": 31.0, "gpqa_diamond": 71.2, "math_500": 91.5, "blended_cost_per_m": 0.18, "composite_intelligence": 74.0, "is_pareto_optimal": true, "is_open_weight": true, "hosted_providers": [ { "provider_id": "google-vertex", "provider_name": "Google Gemini & Vertex AI", "input_cost_per_m": 0.12, "output_cost_per_m": 0.36, "avg_ttft_ms": 115, "tokens_per_sec": 155.0, "is_primary": true }, { "provider_id": "groq", "provider_name": "Groq LPU Inference", "input_cost_per_m": 0.12, "output_cost_per_m": 0.36, "avg_ttft_ms": 75, "tokens_per_sec": 279.0, "is_primary": false } ] }, { "id": "gemini-2-0-flash", "name": "Gemini 2.0 Flash", "organization": "Google DeepMind", "release_date": "2024-12-11", "license": "Proprietary Commercial", "context_window": 1073152, "max_output_tokens": 8192, "input_cost_per_m": 0.1, "output_cost_per_m": 0.4, "cached_input_cost_per_m": 0.025, "swe_bench": 51.8, "mmlu_pro": 78.2, "humaneval": 90.7, "avg_ttft_ms": 280, "tokens_per_sec": 185.0, "ivr_score": 77.7, "wikidata_id": "Q131505676", "official_url": "https://ai.google.dev/pricing", "description": "Previous-generation high-throughput 1M context multimodal workhorse (superseded by Gemini 3.8 Flash and Gemini 3.5 Flash-Lite).", "use_case_summary": "Legacy high-throughput multimodal pipelines migrating to Gemini 3.8 Flash.", "last_updated": "2026-10-16 08:17:46", "cache_write_cost_per_m": 0.1, "architecture": "Dense Multimodal Transformer", "modalities": "Text, Vision, Audio, Video", "parameter_count_b": 35.0, "gpqa_diamond": 63.6, "math_500": 91.1, "blended_cost_per_m": 0.175, "composite_intelligence": 68.6, "is_pareto_optimal": false, "is_open_weight": false, "hosted_providers": [ { "provider_id": "google-vertex", "provider_name": "Google Gemini & Vertex AI", "input_cost_per_m": 0.1, "output_cost_per_m": 0.4, "avg_ttft_ms": 280, "tokens_per_sec": 185.0, "is_primary": true } ] }, { "id": "llama-4-maverick", "name": "Llama 4 Maverick", "organization": "Meta", "release_date": "2025-04-05", "license": "Llama 4 Community License", "context_window": 1048576, "max_output_tokens": 32768, "input_cost_per_m": 0.2, "output_cost_per_m": 0.6, "cached_input_cost_per_m": 0.05, "swe_bench": 62.4, "mmlu_pro": 84.5, "humaneval": 92.6, "avg_ttft_ms": 135, "tokens_per_sec": 175.0, "ivr_score": 77.0, "wikidata_id": "Q131505700", "official_url": "https://llama.meta.com/", "description": "Meta's flagship 128-expert open-weight multimodal MoE model delivering 1M context and rapid LPU/CS-3 inference throughput.", "use_case_summary": "High-throughput open-weight multimodal assistants, enterprise RAG, and low-cost cloud inference on Groq and Cerebras.", "last_updated": "2026-10-16 08:17:46", "cache_write_cost_per_m": 0.2, "architecture": "Multimodal Sparse MoE (17B Active / 400B Total)", "modalities": "Text, Vision", "parameter_count_b": 400.0, "gpqa_diamond": 73.8, "math_500": 91.0, "blended_cost_per_m": 0.3, "composite_intelligence": 75.9, "is_pareto_optimal": false, "is_open_weight": true, "hosted_providers": [ { "provider_id": "google-vertex", "provider_name": "Google Gemini & Vertex AI", "input_cost_per_m": 0.2, "output_cost_per_m": 0.6, "avg_ttft_ms": 150, "tokens_per_sec": 183.8, "is_primary": false }, { "provider_id": "groq", "provider_name": "Groq LPU Inference", "input_cost_per_m": 0.2, "output_cost_per_m": 0.6, "avg_ttft_ms": 80, "tokens_per_sec": 420.0, "is_primary": true }, { "provider_id": "cerebras", "provider_name": "Cerebras CS-3 Wafer-Scale Inference", "input_cost_per_m": 0.2, "output_cost_per_m": 0.6, "avg_ttft_ms": 70, "tokens_per_sec": 560.0, "is_primary": false } ] }, { "id": "glm-5-3-flash", "name": "GLM-5.3-Flash", "organization": "Zhipu AI", "release_date": "2026-02-22", "license": "Open Weights", "context_window": 262144, "max_output_tokens": 16384, "input_cost_per_m": 0.12, "output_cost_per_m": 0.4, "cached_input_cost_per_m": 0.03, "swe_bench": 52.8, "mmlu_pro": 78.4, "humaneval": 89.0, "avg_ttft_ms": 90, "tokens_per_sec": 175.0, "ivr_score": 76.7, "wikidata_id": "Q125934571", "official_url": "https://huggingface.co/zai-org/GLM-5.3-Flash", "description": "Sub-100ms high-throughput open-weight model with 256k context for real-time classification and agent routing.", "use_case_summary": "Sub-100ms ultra-fast inference model designed for real-time customer support, document classification, and agentic workflows.", "last_updated": "2026-10-16 08:17:46", "cache_write_cost_per_m": 0.12, "architecture": "Lightweight High-Throughput Transformer", "modalities": "Text, Code", "parameter_count_b": 16.0, "gpqa_diamond": 63.5, "math_500": 84.0, "blended_cost_per_m": 0.19, "composite_intelligence": 67.6, "is_pareto_optimal": false, "is_open_weight": true, "hosted_providers": [ { "provider_id": "zhipu-ai", "provider_name": "Zhipu AI BigModel Platform", "input_cost_per_m": 0.12, "output_cost_per_m": 0.4, "avg_ttft_ms": 90, "tokens_per_sec": 175.0, "is_primary": true } ] }, { "id": "deepseek-v3", "name": "DeepSeek V3", "organization": "DeepSeek", "release_date": "2024-12-26", "license": "MIT Open Source", "context_window": 131072, "max_output_tokens": 8192, "input_cost_per_m": 0.14, "output_cost_per_m": 0.28, "cached_input_cost_per_m": 0.014, "swe_bench": 42.0, "mmlu_pro": 75.9, "humaneval": 89.1, "avg_ttft_ms": 410, "tokens_per_sec": 78.0, "ivr_score": 75.0, "wikidata_id": "Q131557008", "official_url": "https://api-docs.deepseek.com/quick_start/pricing", "description": "671B parameter Mixture-of-Experts open-weight model (37B active) delivering strong price-to-performance for general text workloads.", "use_case_summary": "Cost-sensitive batch text processing, drafting, and open-weight self-hosted inference.", "last_updated": "2026-10-16 08:17:46", "cache_write_cost_per_m": 0.14, "architecture": "MoE (Multi-head Latent Attention)", "modalities": "Text", "parameter_count_b": 671.0, "gpqa_diamond": 59.1, "math_500": 90.2, "blended_cost_per_m": 0.175, "composite_intelligence": 63.5, "is_pareto_optimal": false, "is_open_weight": true, "hosted_providers": [ { "provider_id": "deepseek", "provider_name": "DeepSeek Platform", "input_cost_per_m": 0.14, "output_cost_per_m": 0.28, "avg_ttft_ms": 410, "tokens_per_sec": 78.0, "is_primary": true }, { "provider_id": "alibaba-cloud", "provider_name": "Alibaba Cloud DashScope", "input_cost_per_m": 0.14, "output_cost_per_m": 0.28, "avg_ttft_ms": 370, "tokens_per_sec": 85.8, "is_primary": false } ] }, { "id": "qwen-3-8-27b", "name": "Qwen 3.8 27B", "organization": "Alibaba Cloud", "release_date": "2026-02-15", "license": "Apache 2.0 Open Source", "context_window": 131072, "max_output_tokens": 16384, "input_cost_per_m": 0.2, "output_cost_per_m": 0.6, "cached_input_cost_per_m": 0.05, "swe_bench": 58.4, "mmlu_pro": 81.2, "humaneval": 91.0, "avg_ttft_ms": 120, "tokens_per_sec": 140.0, "ivr_score": 74.9, "wikidata_id": "Q130489955", "official_url": "https://huggingface.co/Qwen/Qwen3.8-27B", "description": "Compact 27B dense Apache 2.0 model engineered for high-accuracy code generation and single-GPU enterprise hosting.", "use_case_summary": "High-performance 27B open weights optimized for local deployment, code generation, and low-latency production pipelines.", "last_updated": "2026-10-16 08:17:46", "cache_write_cost_per_m": 0.2, "architecture": "Dense Transformer", "modalities": "Text, Code", "parameter_count_b": 27.0, "gpqa_diamond": 68.0, "math_500": 86.4, "blended_cost_per_m": 0.3, "composite_intelligence": 71.6, "is_pareto_optimal": false, "is_open_weight": true, "hosted_providers": [ { "provider_id": "alibaba-cloud", "provider_name": "Alibaba Cloud DashScope", "input_cost_per_m": 0.2, "output_cost_per_m": 0.6, "avg_ttft_ms": 120, "tokens_per_sec": 140.0, "is_primary": true }, { "provider_id": "cerebras", "provider_name": "Cerebras CS-3 Wafer-Scale Inference", "input_cost_per_m": 0.2, "output_cost_per_m": 0.6, "avg_ttft_ms": 70, "tokens_per_sec": 392.0, "is_primary": false } ] }, { "id": "llama-3-3-70b", "name": "Llama 3.3 70B Instruct", "organization": "Meta", "release_date": "2024-12-06", "license": "Llama 3.3 Community License", "context_window": 131072, "max_output_tokens": 32768, "input_cost_per_m": 0.12, "output_cost_per_m": 0.3, "cached_input_cost_per_m": 0.03, "swe_bench": 42.5, "mmlu_pro": 76.8, "humaneval": 88.4, "avg_ttft_ms": 145, "tokens_per_sec": 165.0, "ivr_score": 74.6, "wikidata_id": "Q131505700", "official_url": "https://llama.meta.com/", "description": "Meta's proven 70B dense open-weight model widely deployed across hardware accelerators (Groq, Cerebras) for ultra-fast text generation.", "use_case_summary": "Cost-optimized on-premise deployments, fast text transformation, structured entity extraction, and enterprise dialogue pipelines.", "last_updated": "2026-10-16 08:17:46", "cache_write_cost_per_m": 0.12, "architecture": "Dense Transformer with GQA", "modalities": "Text", "parameter_count_b": 70.0, "gpqa_diamond": 55.6, "math_500": 77.0, "blended_cost_per_m": 0.165, "composite_intelligence": 60.6, "is_pareto_optimal": false, "is_open_weight": true, "hosted_providers": [ { "provider_id": "groq", "provider_name": "Groq LPU Inference", "input_cost_per_m": 0.12, "output_cost_per_m": 0.3, "avg_ttft_ms": 75, "tokens_per_sec": 363.0, "is_primary": true }, { "provider_id": "cerebras", "provider_name": "Cerebras CS-3 Wafer-Scale Inference", "input_cost_per_m": 0.12, "output_cost_per_m": 0.3, "avg_ttft_ms": 60, "tokens_per_sec": 577.5, "is_primary": false } ] }, { "id": "gemini-3-8-flash", "name": "Gemini 3.8 Flash", "organization": "Google", "release_date": "2026-05-12", "license": "Proprietary Commercial", "context_window": 1048576, "max_output_tokens": 65536, "input_cost_per_m": 0.25, "output_cost_per_m": 1.5, "cached_input_cost_per_m": 0.025, "swe_bench": 68.5, "mmlu_pro": 87.4, "humaneval": 94.2, "avg_ttft_ms": 105, "tokens_per_sec": 210.0, "ivr_score": 74.0, "wikidata_id": "Q131505676", "official_url": "https://ai.google.dev/gemini-api/docs/latest-model", "description": "High-velocity 1M-context multimodal reasoning model balancing sub-110ms TTFT latency with deep agentic tool execution and configurable thinking levels.", "use_case_summary": "High-throughput agentic workflows, real-time multimodal RAG, live search grounding, and production code generation at flash unit economics.", "last_updated": "2026-10-16 08:17:46", "cache_write_cost_per_m": 0.25, "architecture": "Hybrid Reasoning Multimodal MoE", "modalities": "Text, Vision, Audio, Video", "parameter_count_b": 85.0, "gpqa_diamond": 79.2, "math_500": 95.1, "blended_cost_per_m": 0.562, "composite_intelligence": 80.7, "is_pareto_optimal": false, "is_open_weight": false, "hosted_providers": [ { "provider_id": "google-vertex", "provider_name": "Google Gemini & Vertex AI", "input_cost_per_m": 0.25, "output_cost_per_m": 1.5, "avg_ttft_ms": 105, "tokens_per_sec": 210.0, "is_primary": true } ] }, { "id": "glm-5-3", "name": "GLM-5.3", "organization": "Zhipu AI", "release_date": "2026-02-18", "license": "Open Weights", "context_window": 131072, "max_output_tokens": 16384, "input_cost_per_m": 0.35, "output_cost_per_m": 1.0, "cached_input_cost_per_m": 0.09, "swe_bench": 56.0, "mmlu_pro": 80.5, "humaneval": 91.5, "avg_ttft_ms": 160, "tokens_per_sec": 120.0, "ivr_score": 69.9, "wikidata_id": "Q125934571", "official_url": "https://huggingface.co/zai-org/GLM-5.3", "description": "Zhipu AI's bilingual flagship foundation model tuned for cross-lingual reasoning, structured function calling, and agentic execution.", "use_case_summary": "Bilingual flagship foundation model optimized for cross-lingual reasoning, long-form structured analysis, and enterprise tool execution.", "last_updated": "2026-10-16 08:17:46", "cache_write_cost_per_m": 0.35, "architecture": "Dense Bilingual Transformer", "modalities": "Text, Code", "parameter_count_b": 65.0, "gpqa_diamond": 67.0, "math_500": 88.0, "blended_cost_per_m": 0.512, "composite_intelligence": 70.7, "is_pareto_optimal": false, "is_open_weight": true, "hosted_providers": [ { "provider_id": "zhipu-ai", "provider_name": "Zhipu AI BigModel Platform", "input_cost_per_m": 0.35, "output_cost_per_m": 1.0, "avg_ttft_ms": 160, "tokens_per_sec": 120.0, "is_primary": true } ] }, { "id": "grok-3-mini", "name": "Grok 3 Mini", "organization": "xAI", "release_date": "2025-02-17", "license": "Proprietary Commercial", "context_window": 131072, "max_output_tokens": 16384, "input_cost_per_m": 0.3, "output_cost_per_m": 1.2, "cached_input_cost_per_m": 0.08, "swe_bench": 54.0, "mmlu_pro": 80.0, "humaneval": 90.0, "avg_ttft_ms": 150, "tokens_per_sec": 160.0, "ivr_score": 69.2, "wikidata_id": "Q123337989", "official_url": "https://x.ai/api", "description": "Fast reasoning model delivering chain-of-thought code and math verification at low token cost.", "use_case_summary": "High-speed reasoning model providing rapid answers and code generation with chain-of-thought capabilities at edge pricing.", "last_updated": "2026-10-16 08:17:46", "cache_write_cost_per_m": 0.3, "architecture": "Compact Reasoning Model", "modalities": "Text, Code", "parameter_count_b": 35.0, "gpqa_diamond": 68.0, "math_500": 87.0, "blended_cost_per_m": 0.525, "composite_intelligence": 69.9, "is_pareto_optimal": false, "is_open_weight": false, "hosted_providers": [ { "provider_id": "xai", "provider_name": "xAI Colossus API", "input_cost_per_m": 0.3, "output_cost_per_m": 1.2, "avg_ttft_ms": 150, "tokens_per_sec": 160.0, "is_primary": true } ] }, { "id": "gpt-4o-mini", "name": "GPT-4o Mini", "organization": "OpenAI", "release_date": "2024-07-18", "license": "Proprietary Commercial", "context_window": 131072, "max_output_tokens": 16384, "input_cost_per_m": 0.15, "output_cost_per_m": 0.6, "cached_input_cost_per_m": 0.075, "swe_bench": 41.0, "mmlu_pro": 68.0, "humaneval": 87.0, "avg_ttft_ms": 220, "tokens_per_sec": 140.0, "ivr_score": 67.2, "wikidata_id": "Q127601736", "official_url": "https://openai.com/api/pricing/", "description": "Low-latency multimodal model optimized for high-volume customer queries, fine-tuning, and structured extraction.", "use_case_summary": "High-volume classification, lightweight RAG chatbots, and structured JSON extraction.", "last_updated": "2026-10-16 08:17:46", "cache_write_cost_per_m": 0.15, "architecture": "Edge-Optimized Omni Model", "modalities": "Text, Vision", "parameter_count_b": 20.0, "gpqa_diamond": 40.2, "math_500": 70.2, "blended_cost_per_m": 0.262, "composite_intelligence": 53.4, "is_pareto_optimal": false, "is_open_weight": false, "hosted_providers": [ { "provider_id": "openai", "provider_name": "OpenAI Platform API", "input_cost_per_m": 0.15, "output_cost_per_m": 0.6, "avg_ttft_ms": 220, "tokens_per_sec": 140.0, "is_primary": true } ] }, { "id": "deepseek-v4-pro", "name": "DeepSeek V4 Pro", "organization": "DeepSeek", "release_date": "2026-09-10", "license": "MIT Open Source", "context_window": 1024000, "max_output_tokens": 128000, "input_cost_per_m": 1.32, "output_cost_per_m": 3.96, "cached_input_cost_per_m": 0.145, "swe_bench": 80.6, "mmlu_pro": 88.4, "humaneval": 94.5, "avg_ttft_ms": 520, "tokens_per_sec": 78.0, "ivr_score": 67.0, "wikidata_id": "Q131557008", "official_url": "https://api-docs.deepseek.com/quick_start/pricing", "description": "DeepSeek's 1.6T parameter open-weight MoE flagship delivering 80.6% SWE-bench Verified and 90.1% GPQA Diamond with 1M context at $1.32/1M input tokens.", "use_case_summary": "Open-weight 1.6T MoE powerhouse designed for cost-effective full-repository code generation, mathematical reasoning, and persistent chain-of-thought multi-step agent workflows.", "last_updated": "2026-10-16 08:17:46", "cache_write_cost_per_m": 1.32, "architecture": "1.6T Sparse MoE with Hybrid CSA/HCA Attention", "modalities": "Text, Vision", "parameter_count_b": 1600.0, "gpqa_diamond": 90.1, "math_500": 97.3, "blended_cost_per_m": 1.98, "composite_intelligence": 87.8, "is_pareto_optimal": false, "is_open_weight": true, "hosted_providers": [ { "provider_id": "deepseek", "provider_name": "DeepSeek Platform", "input_cost_per_m": 1.32, "output_cost_per_m": 3.96, "avg_ttft_ms": 520, "tokens_per_sec": 78.0, "is_primary": true } ] }, { "id": "gemini-2-5-flash", "name": "Gemini 2.5 Flash", "organization": "Google", "release_date": "2025-03-03", "license": "Proprietary Commercial", "context_window": 1048576, "max_output_tokens": 65536, "input_cost_per_m": 0.3, "output_cost_per_m": 2.5, "cached_input_cost_per_m": 0.03, "swe_bench": 58.0, "mmlu_pro": 82.0, "humaneval": 91.5, "avg_ttft_ms": 110, "tokens_per_sec": 180.0, "ivr_score": 66.6, "wikidata_id": "Q131505676", "official_url": "https://ai.google.dev/pricing", "description": "Hybrid reasoning model delivering strong price-performance across 1M context with configurable thinking budgets.", "use_case_summary": "Cost-efficient 1M-context summarization, multimodal extraction, and interactive assistants.", "last_updated": "2026-10-16 08:17:46", "cache_write_cost_per_m": 0.3, "architecture": "Hybrid Reasoning Omni Transformer", "modalities": "Text, Vision, Audio, Video", "parameter_count_b": 70.0, "gpqa_diamond": 70.5, "math_500": 89.0, "blended_cost_per_m": 0.85, "composite_intelligence": 72.7, "is_pareto_optimal": false, "is_open_weight": false, "hosted_providers": [ { "provider_id": "google-vertex", "provider_name": "Google Gemini & Vertex AI", "input_cost_per_m": 0.3, "output_cost_per_m": 2.5, "avg_ttft_ms": 110, "tokens_per_sec": 180.0, "is_primary": true } ] }, { "id": "deepseek-r1", "name": "DeepSeek R1", "organization": "DeepSeek", "release_date": "2025-01-20", "license": "MIT Open Source", "context_window": 131072, "max_output_tokens": 32768, "input_cost_per_m": 0.55, "output_cost_per_m": 2.19, "cached_input_cost_per_m": 0.14, "swe_bench": 49.2, "mmlu_pro": 84.0, "humaneval": 92.8, "avg_ttft_ms": 1450, "tokens_per_sec": 58.0, "ivr_score": 64.8, "wikidata_id": "Q131846875", "official_url": "https://platform.deepseek.com/api-docs/pricing", "description": "MIT-licensed open-weight reasoning powerhouse trained via large-scale reinforcement learning for mathematical and algorithmic synthesis.", "use_case_summary": "Open-weight chain-of-thought reasoning, synthetic data distillation, and complex mathematical verification.", "last_updated": "2026-10-16 08:17:46", "cache_write_cost_per_m": 0.55, "architecture": "MoE Reasoning Transformer", "modalities": "Text", "parameter_count_b": 671.0, "gpqa_diamond": 71.5, "math_500": 97.3, "blended_cost_per_m": 0.96, "composite_intelligence": 72.0, "is_pareto_optimal": false, "is_open_weight": true, "hosted_providers": [ { "provider_id": "deepseek", "provider_name": "DeepSeek Platform", "input_cost_per_m": 0.55, "output_cost_per_m": 2.19, "avg_ttft_ms": 1450, "tokens_per_sec": 58.0, "is_primary": true }, { "provider_id": "alibaba-cloud", "provider_name": "Alibaba Cloud DashScope", "input_cost_per_m": 0.55, "output_cost_per_m": 2.19, "avg_ttft_ms": 1330, "tokens_per_sec": 66.7, "is_primary": false }, { "provider_id": "groq", "provider_name": "Groq LPU Inference", "input_cost_per_m": 0.55, "output_cost_per_m": 2.19, "avg_ttft_ms": 850, "tokens_per_sec": 150.8, "is_primary": false }, { "provider_id": "cerebras", "provider_name": "Cerebras CS-3 Wafer-Scale Inference", "input_cost_per_m": 0.55, "output_cost_per_m": 2.19, "avg_ttft_ms": 700, "tokens_per_sec": 174.0, "is_primary": false } ] }, { "id": "o4-mini", "name": "OpenAI o4-mini", "organization": "OpenAI", "release_date": "2025-04-16", "license": "Proprietary Commercial", "context_window": 204800, "max_output_tokens": 100000, "input_cost_per_m": 1.1, "output_cost_per_m": 4.4, "cached_input_cost_per_m": 0.275, "swe_bench": 68.1, "mmlu_pro": 85.6, "humaneval": 94.0, "avg_ttft_ms": 680, "tokens_per_sec": 125.0, "ivr_score": 63.7, "wikidata_id": "Q130300481", "official_url": "https://openai.com/api/pricing/", "description": "High-throughput multimodal reasoning model delivering frontier STEM and coding accuracy at a fraction of heavyweight reasoning API costs.", "use_case_summary": "Automated PR review, high-volume agentic coding loops, SQL/data analysis, and visual document reasoning.", "last_updated": "2026-10-16 08:17:46", "cache_write_cost_per_m": 1.1, "architecture": "Compact RL Reasoning Transformer", "modalities": "Text, Vision", "parameter_count_b": 55.0, "gpqa_diamond": 81.4, "math_500": 97.8, "blended_cost_per_m": 1.925, "composite_intelligence": 81.1, "is_pareto_optimal": false, "is_open_weight": false, "hosted_providers": [ { "provider_id": "openai", "provider_name": "OpenAI Platform API", "input_cost_per_m": 1.1, "output_cost_per_m": 4.4, "avg_ttft_ms": 680, "tokens_per_sec": 125.0, "is_primary": true } ] }, { "id": "gpt-6-1-sol", "name": "GPT-6.1 Sol", "organization": "OpenAI", "release_date": "2026-10-15", "license": "Proprietary Commercial", "context_window": 1075200, "max_output_tokens": 128000, "input_cost_per_m": 2.0, "output_cost_per_m": 10.0, "cached_input_cost_per_m": 0.1, "swe_bench": 86.4, "mmlu_pro": 91.0, "humaneval": 96.8, "avg_ttft_ms": 520, "tokens_per_sec": 135.0, "ivr_score": 62.9, "wikidata_id": null, "official_url": "https://openai.com/api/pricing", "description": "Delivers near-GPT-6 Astra frontier accuracy at one-fifth the token cost with a 95% cache read discount for high-efficiency agentic coding, computer use, and enterprise workflow automation.", "use_case_summary": "Delivers near-GPT-6 Astra frontier accuracy at one-fifth the token cost with a 95% cache read discount for high-efficiency agentic coding, computer use, and enterprise workflow automation.", "last_updated": "2026-10-16 08:17:46", "cache_write_cost_per_m": 2.0, "architecture": "Adaptive Reasoning Multimodal Transformer", "modalities": "Text, Vision, Audio", "parameter_count_b": null, "gpqa_diamond": 92.4, "math_500": 98.8, "blended_cost_per_m": 4.0, "composite_intelligence": 91.2, "is_pareto_optimal": true, "is_open_weight": false, "hosted_providers": [ { "provider_id": "openai", "provider_name": "OpenAI Platform API", "input_cost_per_m": 2.0, "output_cost_per_m": 10.0, "avg_ttft_ms": 520, "tokens_per_sec": 135.0, "is_primary": true } ] }, { "id": "gemini-4-argon", "name": "Gemini 4 Argon", "organization": "Google", "release_date": "2026-09-15", "license": "Proprietary Commercial", "context_window": 1024000, "max_output_tokens": 1000000, "input_cost_per_m": 2.0, "output_cost_per_m": 10.0, "cached_input_cost_per_m": 0.1, "swe_bench": 84.5, "mmlu_pro": 91.4, "humaneval": 96.3, "avg_ttft_ms": 850, "tokens_per_sec": 115.0, "ivr_score": 62.6, "wikidata_id": "Q131505676", "official_url": "https://ai.google.dev/gemini-api/docs/pricing", "description": "Google DeepMind's flagship Gemini 4 Argon model featuring configurable high-depth reasoning, 1M token native multimodal context, and industry-leading agentic tool synthesis.", "use_case_summary": "Engineered with a 1M output token limit for long-horizon software engineering, cybersecurity vulnerability remediation, and deep multimodal enterprise work across legal and financial domains.", "last_updated": "2026-10-16 08:17:46", "cache_write_cost_per_m": 0.5, "architecture": "Next-Gen Sparse MoE with Adaptive High Thinking", "modalities": "Text, Vision, Audio, Video", "parameter_count_b": 320.0, "gpqa_diamond": 93.8, "math_500": 98.2, "blended_cost_per_m": 4.0, "composite_intelligence": 90.8, "is_pareto_optimal": false, "is_open_weight": false, "hosted_providers": [ { "provider_id": "google-vertex", "provider_name": "Google Gemini & Vertex AI", "input_cost_per_m": 2.0, "output_cost_per_m": 10.0, "avg_ttft_ms": 850, "tokens_per_sec": 115.0, "is_primary": true }, { "provider_id": "cursor-anysphere", "provider_name": "Anysphere Cursor Router", "input_cost_per_m": 2.0, "output_cost_per_m": 10.0, "avg_ttft_ms": 865, "tokens_per_sec": 115.0, "is_primary": false } ] }, { "id": "o3", "name": "OpenAI o3", "organization": "OpenAI", "release_date": "2025-04-16", "license": "Proprietary Commercial", "context_window": 204800, "max_output_tokens": 100000, "input_cost_per_m": 2.0, "output_cost_per_m": 8.0, "cached_input_cost_per_m": 0.5, "swe_bench": 71.7, "mmlu_pro": 89.8, "humaneval": 95.2, "avg_ttft_ms": 1150, "tokens_per_sec": 82.0, "ivr_score": 60.6, "wikidata_id": "Q130300481", "official_url": "https://openai.com/api/pricing/", "description": "OpenAI's deep reasoning specialist achieving state-of-the-art performance on GPQA Diamond, competitive programming, and visual chain-of-thought.", "use_case_summary": "Algorithmic synthesis, mathematical verification, PhD-level scientific research, and multi-step tool planning.", "last_updated": "2026-10-16 08:17:46", "cache_write_cost_per_m": 2.0, "architecture": "Reinforcement Learning Reasoning Transformer", "modalities": "Text, Vision", "parameter_count_b": 300.0, "gpqa_diamond": 87.7, "math_500": 98.6, "blended_cost_per_m": 3.5, "composite_intelligence": 84.8, "is_pareto_optimal": false, "is_open_weight": false, "hosted_providers": [ { "provider_id": "openai", "provider_name": "OpenAI Platform API", "input_cost_per_m": 2.0, "output_cost_per_m": 8.0, "avg_ttft_ms": 1150, "tokens_per_sec": 82.0, "is_primary": true }, { "provider_id": "cursor-anysphere", "provider_name": "Anysphere Cursor Router", "input_cost_per_m": 2.0, "output_cost_per_m": 8.0, "avg_ttft_ms": 1170, "tokens_per_sec": 82.0, "is_primary": false } ] }, { "id": "gemini-3-1-pro", "name": "Gemini 3.1 Pro", "organization": "Google", "release_date": "2026-01-22", "license": "Proprietary Commercial", "context_window": 1048576, "max_output_tokens": 65536, "input_cost_per_m": 1.25, "output_cost_per_m": 10.0, "cached_input_cost_per_m": 0.31, "swe_bench": 71.8, "mmlu_pro": 89.5, "humaneval": 94.8, "avg_ttft_ms": 420, "tokens_per_sec": 98.0, "ivr_score": 60.0, "wikidata_id": "Q131505676", "official_url": "https://ai.google.dev/pricing", "description": "Enterprise-grade 1M-token reasoning model optimized for complex repository analysis, multi-hour video comprehension, and custom tool synthesis.", "use_case_summary": "Complex enterprise architecture planning, multi-file code refactoring, legal/financial corpus auditing, and multimodal research.", "last_updated": "2026-10-16 08:17:46", "cache_write_cost_per_m": 1.25, "architecture": "Deep Reasoning Multimodal MoE", "modalities": "Text, Vision, Audio, Video", "parameter_count_b": 280.0, "gpqa_diamond": 82.4, "math_500": 96.8, "blended_cost_per_m": 3.438, "composite_intelligence": 83.3, "is_pareto_optimal": false, "is_open_weight": false, "hosted_providers": [ { "provider_id": "google-vertex", "provider_name": "Google Gemini & Vertex AI", "input_cost_per_m": 1.25, "output_cost_per_m": 10.0, "avg_ttft_ms": 420, "tokens_per_sec": 98.0, "is_primary": true } ] }, { "id": "o3-mini", "name": "OpenAI o3-mini", "organization": "OpenAI", "release_date": "2025-01-31", "license": "Proprietary Commercial", "context_window": 204800, "max_output_tokens": 100000, "input_cost_per_m": 1.1, "output_cost_per_m": 4.4, "cached_input_cost_per_m": 0.55, "swe_bench": 49.3, "mmlu_pro": 83.2, "humaneval": 92.4, "avg_ttft_ms": 950, "tokens_per_sec": 98.0, "ivr_score": 59.7, "wikidata_id": "Q130300481", "official_url": "https://openai.com/api/pricing/", "description": "Cost-effective text reasoning model purpose-built for STEM logic, theorem validation, and structured code generation.", "use_case_summary": "Competitive math reasoning, automated unit test generation, and structured logic pipelines.", "last_updated": "2026-10-16 08:17:46", "cache_write_cost_per_m": 1.1, "architecture": "Reasoning Transformer", "modalities": "Text", "parameter_count_b": 45.0, "gpqa_diamond": 79.7, "math_500": 97.9, "blended_cost_per_m": 1.925, "composite_intelligence": 73.6, "is_pareto_optimal": false, "is_open_weight": false, "hosted_providers": [ { "provider_id": "openai", "provider_name": "OpenAI Platform API", "input_cost_per_m": 1.1, "output_cost_per_m": 4.4, "avg_ttft_ms": 950, "tokens_per_sec": 98.0, "is_primary": true } ] }, { "id": "gpt-5", "name": "GPT-5", "organization": "OpenAI", "release_date": "2025-11-18", "license": "Proprietary Commercial", "context_window": 409600, "max_output_tokens": 128000, "input_cost_per_m": 2.5, "output_cost_per_m": 10.0, "cached_input_cost_per_m": 0.625, "swe_bench": 74.9, "mmlu_pro": 90.8, "humaneval": 95.8, "avg_ttft_ms": 340, "tokens_per_sec": 105.0, "ivr_score": 59.2, "wikidata_id": "Q125916098", "official_url": "https://openai.com/api/pricing/", "description": "OpenAI's unified flagship system combining deep test-time reasoning with native multimodal perception and 400k context.", "use_case_summary": "Complex enterprise automation, full-stack software engineering, multimodal agent orchestration, and scientific synthesis.", "last_updated": "2026-10-16 08:17:46", "cache_write_cost_per_m": 2.5, "architecture": "Unified Omni-Reasoning MoE", "modalities": "Text, Vision, Audio, Video", "parameter_count_b": 350.0, "gpqa_diamond": 85.4, "math_500": 98.0, "blended_cost_per_m": 4.375, "composite_intelligence": 85.6, "is_pareto_optimal": false, "is_open_weight": false, "hosted_providers": [ { "provider_id": "openai", "provider_name": "OpenAI Platform API", "input_cost_per_m": 2.5, "output_cost_per_m": 10.0, "avg_ttft_ms": 340, "tokens_per_sec": 105.0, "is_primary": true }, { "provider_id": "cursor-anysphere", "provider_name": "Anysphere Cursor Router", "input_cost_per_m": 2.5, "output_cost_per_m": 10.0, "avg_ttft_ms": 360, "tokens_per_sec": 105.0, "is_primary": false } ] }, { "id": "claude-opus-5-5", "name": "Claude Opus 5.5", "organization": "Anthropic", "release_date": "2026-09-18", "license": "Proprietary Commercial", "context_window": 1024000, "max_output_tokens": 128000, "input_cost_per_m": 4.0, "output_cost_per_m": 20.0, "cached_input_cost_per_m": 0.2, "swe_bench": 89.9, "mmlu_pro": 92.1, "humaneval": 97.6, "avg_ttft_ms": 680, "tokens_per_sec": 95.0, "ivr_score": 57.8, "wikidata_id": "Q130268023", "official_url": "https://docs.anthropic.com/en/docs/about-claude/pricing", "description": "Anthropic's September 2026 frontier powerhouse achieving 89.9% SWE-bench Verified and 92.5% GPQA Diamond at a 73% lower token price than previous Opus generations.", "use_case_summary": "Top-ranked frontier intelligence leader for autonomous multi-step agentic coding, repository-scale migrations, and complex scientific reasoning at a 40% lower workload cost than Opus 5.", "last_updated": "2026-10-16 08:17:46", "cache_write_cost_per_m": 5.0, "architecture": "Extended Thinking Hybrid Transformer", "modalities": "Text, Vision", "parameter_count_b": 480.0, "gpqa_diamond": 93.2, "math_500": 98.0, "blended_cost_per_m": 8.0, "composite_intelligence": 92.7, "is_pareto_optimal": true, "is_open_weight": false, "hosted_providers": [ { "provider_id": "anthropic", "provider_name": "Anthropic Claude API", "input_cost_per_m": 4.0, "output_cost_per_m": 20.0, "avg_ttft_ms": 680, "tokens_per_sec": 95.0, "is_primary": true } ] }, { "id": "grok-4", "name": "Grok 4", "organization": "xAI", "release_date": "2026-01-14", "license": "Proprietary Commercial", "context_window": 262144, "max_output_tokens": 65536, "input_cost_per_m": 2.5, "output_cost_per_m": 12.5, "cached_input_cost_per_m": 0.5, "swe_bench": 73.2, "mmlu_pro": 89.9, "humaneval": 95.0, "avg_ttft_ms": 480, "tokens_per_sec": 95.0, "ivr_score": 57.6, "wikidata_id": "Q123337989", "official_url": "https://docs.x.ai/docs/overview", "description": "xAI's flagship reasoning model trained on the 200k-GPU Colossus cluster, excelling at graduate-level science, code synthesis, and live web telemetry.", "use_case_summary": "Real-time financial/market intelligence, PhD-level STEM problem solving, and autonomous code generation.", "last_updated": "2026-10-16 08:17:46", "cache_write_cost_per_m": 2.5, "architecture": "Colossus Sparse MoE with Real-Time Reasoning", "modalities": "Text, Vision, Real-Time Search", "parameter_count_b": 420.0, "gpqa_diamond": 86.2, "math_500": 97.6, "blended_cost_per_m": 5.0, "composite_intelligence": 84.9, "is_pareto_optimal": false, "is_open_weight": false, "hosted_providers": [ { "provider_id": "xai", "provider_name": "xAI Colossus API", "input_cost_per_m": 2.5, "output_cost_per_m": 12.5, "avg_ttft_ms": 480, "tokens_per_sec": 95.0, "is_primary": true }, { "provider_id": "cursor-anysphere", "provider_name": "Anysphere Cursor Router", "input_cost_per_m": 2.5, "output_cost_per_m": 12.5, "avg_ttft_ms": 495, "tokens_per_sec": 95.0, "is_primary": false } ] }, { "id": "gemini-2-5-pro", "name": "Gemini 2.5 Pro", "organization": "Google", "release_date": "2025-03-03", "license": "Proprietary Commercial", "context_window": 1024000, "max_output_tokens": 65536, "input_cost_per_m": 1.25, "output_cost_per_m": 10.0, "cached_input_cost_per_m": 0.3125, "swe_bench": 63.8, "mmlu_pro": 85.2, "humaneval": 89.8, "avg_ttft_ms": 700, "tokens_per_sec": 80.0, "ivr_score": 56.8, "wikidata_id": "Q131505676", "official_url": "https://ai.google.dev/pricing", "description": "Million-token multimodal model capable of analyzing hours of video, audio streams, and massive codebase repositories.", "use_case_summary": "Long-context document and video analysis, large codebase ingestion, and multimodal reasoning.", "last_updated": "2026-10-16 08:17:46", "cache_write_cost_per_m": 1.25, "architecture": "Native Multimodal MoE with Thinking", "modalities": "Text, Vision, Audio, Video", "parameter_count_b": 250.0, "gpqa_diamond": 74.0, "math_500": 95.0, "blended_cost_per_m": 3.438, "composite_intelligence": 77.4, "is_pareto_optimal": false, "is_open_weight": false, "hosted_providers": [ { "provider_id": "google-vertex", "provider_name": "Google Gemini & Vertex AI", "input_cost_per_m": 1.25, "output_cost_per_m": 10.0, "avg_ttft_ms": 700, "tokens_per_sec": 80.0, "is_primary": true } ] }, { "id": "claude-4-5-sonnet", "name": "Claude 4.5 Sonnet", "organization": "Anthropic", "release_date": "2026-04-14", "license": "Proprietary Commercial", "context_window": 204800, "max_output_tokens": 64000, "input_cost_per_m": 3.0, "output_cost_per_m": 15.0, "cached_input_cost_per_m": 0.3, "swe_bench": 75.8, "mmlu_pro": 90.4, "humaneval": 96.0, "avg_ttft_ms": 380, "tokens_per_sec": 84.0, "ivr_score": 56.5, "wikidata_id": "Q132145690", "official_url": "https://www.anthropic.com/pricing", "description": "Anthropic's premier software engineering and autonomous computer-use model featuring interleaved thinking and 90% prompt cache read discounts.", "use_case_summary": "End-to-end autonomous coding agents, complex multi-file debugging, browser/GUI automation, and high-precision technical writing.", "last_updated": "2026-10-16 08:17:46", "cache_write_cost_per_m": 3.75, "architecture": "Extended Thinking Hybrid Transformer", "modalities": "Text, Vision, Computer Use", "parameter_count_b": 220.0, "gpqa_diamond": 84.2, "math_500": 97.5, "blended_cost_per_m": 6.0, "composite_intelligence": 85.5, "is_pareto_optimal": false, "is_open_weight": false, "hosted_providers": [ { "provider_id": "google-vertex", "provider_name": "Google Gemini & Vertex AI", "input_cost_per_m": 3.0, "output_cost_per_m": 15.0, "avg_ttft_ms": 405, "tokens_per_sec": 80.6, "is_primary": false }, { "provider_id": "anthropic", "provider_name": "Anthropic Claude API", "input_cost_per_m": 3.0, "output_cost_per_m": 15.0, "avg_ttft_ms": 380, "tokens_per_sec": 84.0, "is_primary": true }, { "provider_id": "cursor-anysphere", "provider_name": "Anysphere Cursor Router", "input_cost_per_m": 3.0, "output_cost_per_m": 15.0, "avg_ttft_ms": 390, "tokens_per_sec": 84.0, "is_primary": false } ] }, { "id": "claude-3-7-sonnet", "name": "Claude 3.7 Sonnet", "organization": "Anthropic", "release_date": "2025-02-24", "license": "Proprietary Commercial", "context_window": 204800, "max_output_tokens": 64000, "input_cost_per_m": 3.0, "output_cost_per_m": 15.0, "cached_input_cost_per_m": 0.3, "swe_bench": 70.3, "mmlu_pro": 88.0, "humaneval": 93.5, "avg_ttft_ms": 540, "tokens_per_sec": 72.0, "ivr_score": 54.3, "wikidata_id": "Q132145690", "official_url": "https://www.anthropic.com/pricing", "description": "Anthropic's hybrid reasoning model combining instant conversational responses with controllable extended thinking budgets up to 64k tokens.", "use_case_summary": "Full-stack repository development, complex code refactoring, and agentic tool use with predictable prompt cache economics.", "last_updated": "2026-10-16 08:17:46", "cache_write_cost_per_m": 3.75, "architecture": "Hybrid Reasoning Dense Transformer", "modalities": "Text, Vision", "parameter_count_b": 200.0, "gpqa_diamond": 78.5, "math_500": 96.2, "blended_cost_per_m": 6.0, "composite_intelligence": 81.5, "is_pareto_optimal": false, "is_open_weight": false, "hosted_providers": [ { "provider_id": "anthropic", "provider_name": "Anthropic Claude API", "input_cost_per_m": 3.0, "output_cost_per_m": 15.0, "avg_ttft_ms": 540, "tokens_per_sec": 72.0, "is_primary": true }, { "provider_id": "cursor-anysphere", "provider_name": "Anysphere Cursor Router", "input_cost_per_m": 3.0, "output_cost_per_m": 15.0, "avg_ttft_ms": 555, "tokens_per_sec": 72.0, "is_primary": false } ] }, { "id": "grok-3", "name": "Grok 3", "organization": "xAI", "release_date": "2025-02-17", "license": "Proprietary Commercial", "context_window": 134144, "max_output_tokens": 16384, "input_cost_per_m": 3.0, "output_cost_per_m": 15.0, "cached_input_cost_per_m": 0.75, "swe_bench": 65.8, "mmlu_pro": 80.0, "humaneval": 91.5, "avg_ttft_ms": 620, "tokens_per_sec": 88.0, "ivr_score": 52.7, "wikidata_id": "Q123337989", "official_url": "https://docs.x.ai/docs/overview", "description": "Large-scale xAI foundation model specialized in scientific reasoning, real-time data analysis, and enterprise logic.", "use_case_summary": "Massive compute foundation model specialized in PhD-level science reasoning, real-time data analysis, and enterprise logic.", "last_updated": "2026-10-16 08:17:46", "cache_write_cost_per_m": 3.0, "architecture": "MoE Transformer", "modalities": "Text, Vision", "parameter_count_b": 350.0, "gpqa_diamond": 84.6, "math_500": 89.8, "blended_cost_per_m": 6.0, "composite_intelligence": 77.9, "is_pareto_optimal": false, "is_open_weight": false, "hosted_providers": [ { "provider_id": "xai", "provider_name": "xAI Colossus API", "input_cost_per_m": 3.0, "output_cost_per_m": 15.0, "avg_ttft_ms": 620, "tokens_per_sec": 88.0, "is_primary": true } ] }, { "id": "claude-3-5-haiku", "name": "Claude 3.5 Haiku", "organization": "Anthropic", "release_date": "2024-11-04", "license": "Proprietary Commercial", "context_window": 204800, "max_output_tokens": 8192, "input_cost_per_m": 0.8, "output_cost_per_m": 4.0, "cached_input_cost_per_m": 0.08, "swe_bench": 40.6, "mmlu_pro": 71.0, "humaneval": 88.5, "avg_ttft_ms": 190, "tokens_per_sec": 130.0, "ivr_score": 52.4, "wikidata_id": "Q126848243", "official_url": "https://www.anthropic.com/pricing", "description": "Fast, compact Anthropic model delivering sub-200ms responses with 90% prompt caching discounts for interactive applications.", "use_case_summary": "Rapid customer-facing chat, sub-agent routing, and real-time code autocomplete.", "last_updated": "2026-10-16 08:17:46", "cache_write_cost_per_m": 1.0, "architecture": "High-Speed Compact Transformer", "modalities": "Text, Vision", "parameter_count_b": 30.0, "gpqa_diamond": 41.6, "math_500": 69.4, "blended_cost_per_m": 1.6, "composite_intelligence": 54.2, "is_pareto_optimal": false, "is_open_weight": false, "hosted_providers": [ { "provider_id": "anthropic", "provider_name": "Anthropic Claude API", "input_cost_per_m": 0.8, "output_cost_per_m": 4.0, "avg_ttft_ms": 190, "tokens_per_sec": 130.0, "is_primary": true } ] }, { "id": "mistral-large-2411", "name": "Mistral Large 2 (2411)", "organization": "Mistral AI", "release_date": "2024-11-18", "license": "Mistral Commercial License", "context_window": 131072, "max_output_tokens": 8192, "input_cost_per_m": 2.0, "output_cost_per_m": 6.0, "cached_input_cost_per_m": 0.2, "swe_bench": 45.2, "mmlu_pro": 74.8, "humaneval": 88.5, "avg_ttft_ms": 380, "tokens_per_sec": 75.0, "ivr_score": 50.1, "wikidata_id": "Q125633842", "official_url": "https://docs.mistral.ai/getting-started/models/", "description": "European data-sovereign 123B enterprise flagship tuned for multilingual accuracy, strict JSON schemas, and EU regulatory compliance.", "use_case_summary": "Sovereign enterprise multilingual flagship tuned for reasoning, strict JSON schemas, and European regulatory compliance.", "last_updated": "2026-10-16 08:17:46", "cache_write_cost_per_m": 2.0, "architecture": "Dense Transformer", "modalities": "Text", "parameter_count_b": 123.0, "gpqa_diamond": 54.2, "math_500": 76.5, "blended_cost_per_m": 3.0, "composite_intelligence": 60.7, "is_pareto_optimal": false, "is_open_weight": false, "hosted_providers": [] }, { "id": "gpt-4o", "name": "GPT-4o", "organization": "OpenAI", "release_date": "2024-05-13", "license": "Proprietary Commercial", "context_window": 131072, "max_output_tokens": 16384, "input_cost_per_m": 2.5, "output_cost_per_m": 10.0, "cached_input_cost_per_m": 1.25, "swe_bench": 48.0, "mmlu_pro": 75.9, "humaneval": 90.2, "avg_ttft_ms": 280, "tokens_per_sec": 100.0, "ivr_score": 47.6, "wikidata_id": "Q125916098", "official_url": "https://openai.com/api/pricing/", "description": "Multimodal enterprise baseline combining native text, vision, and audio processing for real-time applications.", "use_case_summary": "General enterprise assistants, multimodal customer support, and structured JSON output pipelines.", "last_updated": "2026-10-16 08:17:46", "cache_write_cost_per_m": 2.5, "architecture": "Omni-Modal Transformer", "modalities": "Text, Vision, Audio", "parameter_count_b": 200.0, "gpqa_diamond": 53.6, "math_500": 76.6, "blended_cost_per_m": 4.375, "composite_intelligence": 61.8, "is_pareto_optimal": false, "is_open_weight": false, "hosted_providers": [ { "provider_id": "openai", "provider_name": "OpenAI Platform API", "input_cost_per_m": 2.5, "output_cost_per_m": 10.0, "avg_ttft_ms": 280, "tokens_per_sec": 100.0, "is_primary": true } ] }, { "id": "gpt-6-astra", "name": "GPT-6 Astra", "organization": "OpenAI", "release_date": "2026-09-22", "license": "Proprietary Commercial", "context_window": 1075200, "max_output_tokens": 128000, "input_cost_per_m": 10.0, "output_cost_per_m": 50.0, "cached_input_cost_per_m": 1.0, "swe_bench": 75.7, "mmlu_pro": 91.2, "humaneval": 96.8, "avg_ttft_ms": 740, "tokens_per_sec": 74.0, "ivr_score": 47.6, "wikidata_id": "Q131760000", "official_url": "https://openai.com/api/pricing/", "description": "OpenAI's September 2026 flagship reasoning and computer-use foundation model scoring 96.0% on GPQA Diamond with a 1.05M token context window.", "use_case_summary": "High-autonomy flagship reasoning model engineered for end-to-end computer and browser use, advanced scientific research, and complex multi-turn software engineering.", "last_updated": "2026-10-16 08:17:46", "cache_write_cost_per_m": 12.5, "architecture": "Unified Frontier Reasoning Transformer", "modalities": "Text, Vision, Audio", "parameter_count_b": 600.0, "gpqa_diamond": 96.0, "math_500": 98.0, "blended_cost_per_m": 20.0, "composite_intelligence": 88.1, "is_pareto_optimal": false, "is_open_weight": false, "hosted_providers": [ { "provider_id": "openai", "provider_name": "OpenAI Platform API", "input_cost_per_m": 10.0, "output_cost_per_m": 50.0, "avg_ttft_ms": 740, "tokens_per_sec": 74.0, "is_primary": true } ] }, { "id": "claude-3-5-sonnet", "name": "Claude 3.5 Sonnet", "organization": "Anthropic", "release_date": "2024-06-20", "license": "Proprietary Commercial", "context_window": 204800, "max_output_tokens": 8192, "input_cost_per_m": 3.0, "output_cost_per_m": 15.0, "cached_input_cost_per_m": 0.3, "swe_bench": 49.0, "mmlu_pro": 78.0, "humaneval": 92.0, "avg_ttft_ms": 310, "tokens_per_sec": 80.0, "ivr_score": 46.2, "wikidata_id": "Q126848243", "official_url": "https://www.anthropic.com/pricing", "description": "Proven enterprise workhorse for software engineering, visual chart extraction, and nuanced prose generation.", "use_case_summary": "Legacy enterprise workflows, code review pipelines, and human-like content generation.", "last_updated": "2026-10-16 08:17:46", "cache_write_cost_per_m": 3.75, "architecture": "Frontier Multimodal Transformer", "modalities": "Text, Vision", "parameter_count_b": 175.0, "gpqa_diamond": 59.4, "math_500": 78.3, "blended_cost_per_m": 6.0, "composite_intelligence": 64.2, "is_pareto_optimal": false, "is_open_weight": false, "hosted_providers": [ { "provider_id": "anthropic", "provider_name": "Anthropic Claude API", "input_cost_per_m": 3.0, "output_cost_per_m": 15.0, "avg_ttft_ms": 310, "tokens_per_sec": 80.0, "is_primary": true } ] }, { "id": "claude-4-opus", "name": "Claude 4 Opus", "organization": "Anthropic", "release_date": "2026-03-20", "license": "Proprietary Commercial", "context_window": 204800, "max_output_tokens": 64000, "input_cost_per_m": 15.0, "output_cost_per_m": 75.0, "cached_input_cost_per_m": 1.5, "swe_bench": 77.2, "mmlu_pro": 91.8, "humaneval": 96.8, "avg_ttft_ms": 920, "tokens_per_sec": 46.0, "ivr_score": 43.6, "wikidata_id": "Q132145690", "official_url": "https://www.anthropic.com/pricing", "description": "Anthropic's ultra-dense flagship model built for multi-hour autonomous agent horizons, novel scientific synthesis, and zero-defect systems architecture.", "use_case_summary": "Mission-critical R&D, long-horizon autonomous coding tasks, complex theorem proving, and executive strategic synthesis.", "last_updated": "2026-10-16 08:17:46", "cache_write_cost_per_m": 18.75, "architecture": "Frontier Deep Reasoning Transformer", "modalities": "Text, Vision", "parameter_count_b": 400.0, "gpqa_diamond": 87.1, "math_500": 98.1, "blended_cost_per_m": 30.0, "composite_intelligence": 87.0, "is_pareto_optimal": false, "is_open_weight": false, "hosted_providers": [ { "provider_id": "anthropic", "provider_name": "Anthropic Claude API", "input_cost_per_m": 15.0, "output_cost_per_m": 75.0, "avg_ttft_ms": 920, "tokens_per_sec": 46.0, "is_primary": true } ] }, { "id": "o1", "name": "o1", "organization": "OpenAI", "release_date": "2024-12-05", "license": "Proprietary Commercial", "context_window": 204800, "max_output_tokens": 100000, "input_cost_per_m": 15.0, "output_cost_per_m": 60.0, "cached_input_cost_per_m": 7.5, "swe_bench": 48.9, "mmlu_pro": 83.3, "humaneval": 92.4, "avg_ttft_ms": 2600, "tokens_per_sec": 48.0, "ivr_score": 37.2, "wikidata_id": "Q130300481", "official_url": "https://openai.com/api/pricing/", "description": "First-generation heavyweight reasoning model engineered for academic research, complex mathematics, and deep multimodal analysis.", "use_case_summary": "Legacy reasoning benchmarks, scientific literature review, and complex mathematical proofs.", "last_updated": "2026-10-16 08:17:46", "cache_write_cost_per_m": 15.0, "architecture": "Reasoning RL Transformer", "modalities": "Text, Vision", "parameter_count_b": 300.0, "gpqa_diamond": 78.0, "math_500": 96.4, "blended_cost_per_m": 26.25, "composite_intelligence": 72.8, "is_pareto_optimal": false, "is_open_weight": false, "hosted_providers": [ { "provider_id": "openai", "provider_name": "OpenAI Platform API", "input_cost_per_m": 15.0, "output_cost_per_m": 60.0, "avg_ttft_ms": 2600, "tokens_per_sec": 48.0, "is_primary": true } ] } ] }