{
  "schema_version": "2026.11.1",
  "generated_at": "2026-11-21T23:13:33.380Z",
  "source": "https://tokenledger.dev",
  "public_rate_limit_cap": "60 requests/hour per IP (Pro Feed: 10,000 pulls/month via Stripe via Grok Bot)",
  "models": [
    {
      "model_key": "google/gemini-4-argon",
      "slug": "gemini-4-argon",
      "name": "Gemini 4 Argon",
      "provider": "Google DeepMind",
      "provider_slug": "google-deepmind",
      "tier": "Fast Frontier Runtime",
      "license": "Proprietary API",
      "release_date": "2026-09",
      "context_window": 2000000,
      "max_output_tokens": 65536,
      "pricing": {
        "input_per_1m_usd": 0.25,
        "output_per_1m_usd": 1,
        "cache_write_multiplier": 1.15,
        "cache_read_multiplier": 0.2,
        "batch_discount_multiplier": 0.5,
        "long_context_threshold_tokens": 250000,
        "long_context_input_multiplier": 2,
        "long_context_output_multiplier": 1.5,
        "eu_residency_markup": 1.05
      },
      "tokenomics": {
        "tokenizer_bloat_factor": 1.02,
        "default_reasoning_multiplier": 1.25,
        "supports_reasoning_effort": true,
        "supports_prompt_caching": true,
        "supports_batch_api": true
      },
      "performance": {
        "ttft_ms": 210,
        "output_tps": 195,
        "coding_index": 89.4,
        "reasoning_index": 88.1,
        "agentic_tool_score": 91.2,
        "uptime_30d_pct": 99.95
      },
      "use_cases": [
        "high-volume-rag-extraction",
        "realtime-voice-chat",
        "long-context-codebase-audit",
        "autonomous-coding-agents"
      ],
      "affiliate_partner_slug": "vertex-cloud-credits",
      "summary": "High-speed multimodal frontier runtime with 2M context window, sub-220ms TTFT, and exceptional throughput-to-cost efficiency for production RAG and agent routing.",
      "derived_finops": {
        "cache_break_even_ratio": 0.1579,
        "benchmark_10k_agent_tasks_usd": 192.95
      }
    },
    {
      "model_key": "google/gemini-4-carbon",
      "slug": "gemini-4-carbon",
      "name": "Gemini 4 Carbon",
      "provider": "Google DeepMind",
      "provider_slug": "google-deepmind",
      "tier": "Heavy Reasoning & Build",
      "license": "Proprietary API",
      "release_date": "2026-09",
      "context_window": 2000000,
      "max_output_tokens": 65536,
      "pricing": {
        "input_per_1m_usd": 2,
        "output_per_1m_usd": 10,
        "cache_write_multiplier": 1.2,
        "cache_read_multiplier": 0.15,
        "batch_discount_multiplier": 0.5,
        "long_context_threshold_tokens": 200000,
        "long_context_input_multiplier": 2,
        "long_context_output_multiplier": 1.5,
        "eu_residency_markup": 1.05
      },
      "tokenomics": {
        "tokenizer_bloat_factor": 1.02,
        "default_reasoning_multiplier": 2.2,
        "supports_reasoning_effort": true,
        "supports_prompt_caching": true,
        "supports_batch_api": true
      },
      "performance": {
        "ttft_ms": 540,
        "output_tps": 92,
        "coding_index": 96.2,
        "reasoning_index": 97.4,
        "agentic_tool_score": 96.8,
        "uptime_30d_pct": 99.93
      },
      "use_cases": [
        "autonomous-coding-agents",
        "deep-reasoning-tribunal",
        "long-context-codebase-audit"
      ],
      "affiliate_partner_slug": "vertex-cloud-credits",
      "summary": "Flagship heavy reasoning and architectural build model combining a 2M context window with top-tier multi-file software engineering and deep system verification.",
      "derived_finops": {
        "cache_break_even_ratio": 0.1905,
        "benchmark_10k_agent_tasks_usd": 2058.05
      }
    },
    {
      "model_key": "anthropic/claude-opus-5.5",
      "slug": "claude-opus-5-5",
      "name": "Claude Opus 5.5",
      "provider": "Anthropic",
      "provider_slug": "anthropic",
      "tier": "Heavy Reasoning & Build",
      "license": "Proprietary API",
      "release_date": "2026-08",
      "context_window": 500000,
      "max_output_tokens": 32768,
      "pricing": {
        "input_per_1m_usd": 5,
        "output_per_1m_usd": 25,
        "cache_write_multiplier": 1.25,
        "cache_read_multiplier": 0.1,
        "batch_discount_multiplier": 0.5,
        "long_context_threshold_tokens": 200000,
        "long_context_input_multiplier": 2,
        "long_context_output_multiplier": 1.5,
        "eu_residency_markup": 1.1
      },
      "tokenomics": {
        "tokenizer_bloat_factor": 1.18,
        "default_reasoning_multiplier": 2.4,
        "supports_reasoning_effort": true,
        "supports_prompt_caching": true,
        "supports_batch_api": true
      },
      "performance": {
        "ttft_ms": 680,
        "output_tps": 64,
        "coding_index": 97.1,
        "reasoning_index": 96.9,
        "agentic_tool_score": 97.5,
        "uptime_30d_pct": 99.89
      },
      "use_cases": [
        "autonomous-coding-agents",
        "deep-reasoning-tribunal"
      ],
      "affiliate_partner_slug": "bedrock-gateway-pack",
      "summary": "Ultra-capable agentic software engineering and complex research model; requires aggressive prompt-cache hits (90% discount) to tame high output and cache-write rates.",
      "derived_finops": {
        "cache_break_even_ratio": 0.2174,
        "benchmark_10k_agent_tasks_usd": 6112.7
      }
    },
    {
      "model_key": "anthropic/claude-sonnet-5.5",
      "slug": "claude-sonnet-5-5",
      "name": "Claude Sonnet 5.5",
      "provider": "Anthropic",
      "provider_slug": "anthropic",
      "tier": "Balanced Production",
      "license": "Proprietary API",
      "release_date": "2026-08",
      "context_window": 500000,
      "max_output_tokens": 32768,
      "pricing": {
        "input_per_1m_usd": 3,
        "output_per_1m_usd": 15,
        "cache_write_multiplier": 1.25,
        "cache_read_multiplier": 0.1,
        "batch_discount_multiplier": 0.5,
        "long_context_threshold_tokens": 200000,
        "long_context_input_multiplier": 2,
        "long_context_output_multiplier": 1.5,
        "eu_residency_markup": 1.1
      },
      "tokenomics": {
        "tokenizer_bloat_factor": 1.18,
        "default_reasoning_multiplier": 1.75,
        "supports_reasoning_effort": true,
        "supports_prompt_caching": true,
        "supports_batch_api": true
      },
      "performance": {
        "ttft_ms": 430,
        "output_tps": 98,
        "coding_index": 95.4,
        "reasoning_index": 93.8,
        "agentic_tool_score": 95.9,
        "uptime_30d_pct": 99.92
      },
      "use_cases": [
        "autonomous-coding-agents",
        "long-context-codebase-audit"
      ],
      "affiliate_partner_slug": "bedrock-gateway-pack",
      "summary": "Industry workhorse for IDE coding agents and tool-calling workflows, balancing 95+ coding accuracy with strong 0.10x prompt-cache read economics.",
      "derived_finops": {
        "cache_break_even_ratio": 0.2174,
        "benchmark_10k_agent_tasks_usd": 3149.89
      }
    },
    {
      "model_key": "anthropic/claude-haiku-5.5",
      "slug": "claude-haiku-5-5",
      "name": "Claude Haiku 5.5",
      "provider": "Anthropic",
      "provider_slug": "anthropic",
      "tier": "Fast Frontier Runtime",
      "license": "Proprietary API",
      "release_date": "2026-09",
      "context_window": 256000,
      "max_output_tokens": 16384,
      "pricing": {
        "input_per_1m_usd": 0.8,
        "output_per_1m_usd": 4,
        "cache_write_multiplier": 1.25,
        "cache_read_multiplier": 0.1,
        "batch_discount_multiplier": 0.5,
        "long_context_threshold_tokens": 200000,
        "long_context_input_multiplier": 1.5,
        "long_context_output_multiplier": 1.25,
        "eu_residency_markup": 1.1
      },
      "tokenomics": {
        "tokenizer_bloat_factor": 1.18,
        "default_reasoning_multiplier": 1.15,
        "supports_reasoning_effort": false,
        "supports_prompt_caching": true,
        "supports_batch_api": true
      },
      "performance": {
        "ttft_ms": 240,
        "output_tps": 168,
        "coding_index": 88.2,
        "reasoning_index": 85.6,
        "agentic_tool_score": 89,
        "uptime_30d_pct": 99.96
      },
      "use_cases": [
        "high-volume-rag-extraction",
        "realtime-voice-chat"
      ],
      "affiliate_partner_slug": "bedrock-gateway-pack",
      "summary": "Low-latency sub-agent and classification specialist with crisp instruction adherence and 90% cached system-prompt savings.",
      "derived_finops": {
        "cache_break_even_ratio": 0.2174,
        "benchmark_10k_agent_tasks_usd": 680.67
      }
    },
    {
      "model_key": "xai/grok-4-7-heavy",
      "slug": "grok-4-7-heavy",
      "name": "Grok 4.7 Heavy",
      "provider": "xAI",
      "provider_slug": "xai",
      "tier": "Heavy Reasoning & Build",
      "license": "Proprietary API",
      "release_date": "2026-10",
      "context_window": 1000000,
      "max_output_tokens": 65536,
      "pricing": {
        "input_per_1m_usd": 3.5,
        "output_per_1m_usd": 16,
        "cache_write_multiplier": 1,
        "cache_read_multiplier": 0.25,
        "batch_discount_multiplier": 0.5,
        "long_context_threshold_tokens": 256000,
        "long_context_input_multiplier": 1.75,
        "long_context_output_multiplier": 1.5,
        "eu_residency_markup": 1.12
      },
      "tokenomics": {
        "tokenizer_bloat_factor": 1.05,
        "default_reasoning_multiplier": 2.6,
        "supports_reasoning_effort": true,
        "supports_prompt_caching": true,
        "supports_batch_api": true
      },
      "performance": {
        "ttft_ms": 590,
        "output_tps": 86,
        "coding_index": 95.8,
        "reasoning_index": 98.1,
        "agentic_tool_score": 96.4,
        "uptime_30d_pct": 99.91
      },
      "use_cases": [
        "deep-reasoning-tribunal",
        "autonomous-coding-agents"
      ],
      "affiliate_partner_slug": "ai-gateway-pro",
      "summary": "Adversarial multi-perspective tribunal and deep verification engine with zero cache-write penalty (1.0x) and top-ranked 98.1 reasoning score.",
      "derived_finops": {
        "cache_break_even_ratio": 0,
        "benchmark_10k_agent_tasks_usd": 3886.71
      }
    },
    {
      "model_key": "openai/gpt-5-2",
      "slug": "gpt-5-2",
      "name": "GPT-5.2",
      "provider": "OpenAI",
      "provider_slug": "openai",
      "tier": "Balanced Production",
      "license": "Proprietary API",
      "release_date": "2026-07",
      "context_window": 400000,
      "max_output_tokens": 32768,
      "pricing": {
        "input_per_1m_usd": 2.25,
        "output_per_1m_usd": 9.5,
        "cache_write_multiplier": 1,
        "cache_read_multiplier": 0.25,
        "batch_discount_multiplier": 0.5,
        "long_context_threshold_tokens": 200000,
        "long_context_input_multiplier": 2,
        "long_context_output_multiplier": 1.5,
        "eu_residency_markup": 1.1
      },
      "tokenomics": {
        "tokenizer_bloat_factor": 1,
        "default_reasoning_multiplier": 1.85,
        "supports_reasoning_effort": true,
        "supports_prompt_caching": true,
        "supports_batch_api": true
      },
      "performance": {
        "ttft_ms": 390,
        "output_tps": 108,
        "coding_index": 94.9,
        "reasoning_index": 95.2,
        "agentic_tool_score": 95.5,
        "uptime_30d_pct": 99.91
      },
      "use_cases": [
        "autonomous-coding-agents",
        "high-volume-rag-extraction"
      ],
      "affiliate_partner_slug": "ai-gateway-pro",
      "summary": "Generalist production flagship featuring automatic zero-surcharge prompt caching (75% read discount) and compact o200k-next tokenization.",
      "derived_finops": {
        "cache_break_even_ratio": 0,
        "benchmark_10k_agent_tasks_usd": 1967.06
      }
    },
    {
      "model_key": "openai/gpt-5-2-mini",
      "slug": "gpt-5-2-mini",
      "name": "GPT-5.2 Mini",
      "provider": "OpenAI",
      "provider_slug": "openai",
      "tier": "Fast Frontier Runtime",
      "license": "Proprietary API",
      "release_date": "2026-07",
      "context_window": 400000,
      "max_output_tokens": 32768,
      "pricing": {
        "input_per_1m_usd": 0.2,
        "output_per_1m_usd": 0.85,
        "cache_write_multiplier": 1,
        "cache_read_multiplier": 0.25,
        "batch_discount_multiplier": 0.5,
        "long_context_threshold_tokens": 200000,
        "long_context_input_multiplier": 1.5,
        "long_context_output_multiplier": 1.25,
        "eu_residency_markup": 1.1
      },
      "tokenomics": {
        "tokenizer_bloat_factor": 1,
        "default_reasoning_multiplier": 1.3,
        "supports_reasoning_effort": true,
        "supports_prompt_caching": true,
        "supports_batch_api": true
      },
      "performance": {
        "ttft_ms": 225,
        "output_tps": 182,
        "coding_index": 87.6,
        "reasoning_index": 86.9,
        "agentic_tool_score": 88.4,
        "uptime_30d_pct": 99.95
      },
      "use_cases": [
        "high-volume-rag-extraction",
        "realtime-voice-chat"
      ],
      "affiliate_partner_slug": "ai-gateway-pro",
      "summary": "Cost-optimized structured JSON extraction and customer-facing chat runtime with zero cache-write overhead.",
      "derived_finops": {
        "cache_break_even_ratio": 0,
        "benchmark_10k_agent_tasks_usd": 154.28
      }
    },
    {
      "model_key": "openai/o4-pro",
      "slug": "openai-o4-pro",
      "name": "OpenAI o4-pro",
      "provider": "OpenAI",
      "provider_slug": "openai",
      "tier": "Heavy Reasoning & Build",
      "license": "Proprietary API",
      "release_date": "2026-06",
      "context_window": 256000,
      "max_output_tokens": 100000,
      "pricing": {
        "input_per_1m_usd": 8,
        "output_per_1m_usd": 32,
        "cache_write_multiplier": 1,
        "cache_read_multiplier": 0.5,
        "batch_discount_multiplier": 0.5,
        "long_context_threshold_tokens": 128000,
        "long_context_input_multiplier": 2,
        "long_context_output_multiplier": 2,
        "eu_residency_markup": 1.15
      },
      "tokenomics": {
        "tokenizer_bloat_factor": 1,
        "default_reasoning_multiplier": 3.8,
        "supports_reasoning_effort": true,
        "supports_prompt_caching": true,
        "supports_batch_api": true
      },
      "performance": {
        "ttft_ms": 1450,
        "output_tps": 52,
        "coding_index": 96.5,
        "reasoning_index": 98.4,
        "agentic_tool_score": 94.1,
        "uptime_30d_pct": 99.84
      },
      "use_cases": [
        "deep-reasoning-tribunal"
      ],
      "affiliate_partner_slug": "ai-gateway-pro",
      "summary": "Extreme formal theorem, algorithmic synthesis, and math verification engine — watch out for 3.8x hidden chain-of-thought output token inflation.",
      "derived_finops": {
        "cache_break_even_ratio": 0,
        "benchmark_10k_agent_tasks_usd": 10900
      }
    },
    {
      "model_key": "deepseek/deepseek-v3-2",
      "slug": "deepseek-v3-2",
      "name": "DeepSeek V3.2",
      "provider": "DeepSeek",
      "provider_slug": "deepseek",
      "tier": "Open-Weight Efficiency",
      "license": "MIT Open-Weight",
      "release_date": "2026-08",
      "context_window": 164000,
      "max_output_tokens": 16384,
      "pricing": {
        "input_per_1m_usd": 0.14,
        "output_per_1m_usd": 0.42,
        "cache_write_multiplier": 1,
        "cache_read_multiplier": 0.15,
        "batch_discount_multiplier": 0.65,
        "long_context_threshold_tokens": 128000,
        "long_context_input_multiplier": 1.5,
        "long_context_output_multiplier": 1.5,
        "eu_residency_markup": 1.25
      },
      "tokenomics": {
        "tokenizer_bloat_factor": 1.06,
        "default_reasoning_multiplier": 1,
        "supports_reasoning_effort": false,
        "supports_prompt_caching": true,
        "supports_batch_api": true
      },
      "performance": {
        "ttft_ms": 340,
        "output_tps": 118,
        "coding_index": 90.8,
        "reasoning_index": 89.2,
        "agentic_tool_score": 88.7,
        "uptime_30d_pct": 99.6
      },
      "use_cases": [
        "high-volume-rag-extraction",
        "eu-sovereign-self-hosted",
        "autonomous-coding-agents"
      ],
      "affiliate_partner_slug": "together-gpu-cloud",
      "summary": "Unbeatable raw $/1M token MoE workhorse; deployable on sovereign EU GPU clusters via Together/RunPod to eliminate third-party data-residency concerns.",
      "derived_finops": {
        "cache_break_even_ratio": 0,
        "benchmark_10k_agent_tasks_usd": 88.36
      }
    },
    {
      "model_key": "deepseek/deepseek-r2",
      "slug": "deepseek-r2",
      "name": "DeepSeek R2",
      "provider": "DeepSeek",
      "provider_slug": "deepseek",
      "tier": "Open-Weight Efficiency",
      "license": "MIT Open-Weight",
      "release_date": "2026-09",
      "context_window": 164000,
      "max_output_tokens": 32768,
      "pricing": {
        "input_per_1m_usd": 0.45,
        "output_per_1m_usd": 1.85,
        "cache_write_multiplier": 1,
        "cache_read_multiplier": 0.15,
        "batch_discount_multiplier": 0.6,
        "long_context_threshold_tokens": 128000,
        "long_context_input_multiplier": 1.5,
        "long_context_output_multiplier": 1.5,
        "eu_residency_markup": 1.25
      },
      "tokenomics": {
        "tokenizer_bloat_factor": 1.06,
        "default_reasoning_multiplier": 2.9,
        "supports_reasoning_effort": true,
        "supports_prompt_caching": true,
        "supports_batch_api": true
      },
      "performance": {
        "ttft_ms": 620,
        "output_tps": 84,
        "coding_index": 94.1,
        "reasoning_index": 95.7,
        "agentic_tool_score": 91.5,
        "uptime_30d_pct": 99.65
      },
      "use_cases": [
        "deep-reasoning-tribunal",
        "eu-sovereign-self-hosted",
        "autonomous-coding-agents"
      ],
      "affiliate_partner_slug": "together-gpu-cloud",
      "summary": "Open-weight reasoning titan delivering 95+ reasoning accuracy at ~80% lower headline output cost than proprietary heavyweights, even after 2.9x CoT inflation.",
      "derived_finops": {
        "cache_break_even_ratio": 0,
        "benchmark_10k_agent_tasks_usd": 475.55
      }
    },
    {
      "model_key": "meta/llama-4-maverick-400b",
      "slug": "llama-4-maverick-400b",
      "name": "Llama 4 Maverick 400B",
      "provider": "Meta (Hosted Cloud)",
      "provider_slug": "meta",
      "tier": "Open-Weight Efficiency",
      "license": "Llama 4 Community",
      "release_date": "2026-05",
      "context_window": 1000000,
      "max_output_tokens": 32768,
      "pricing": {
        "input_per_1m_usd": 0.28,
        "output_per_1m_usd": 0.88,
        "cache_write_multiplier": 1,
        "cache_read_multiplier": 0.5,
        "batch_discount_multiplier": 0.5,
        "long_context_threshold_tokens": 256000,
        "long_context_input_multiplier": 1.5,
        "long_context_output_multiplier": 1.5,
        "eu_residency_markup": 1.08
      },
      "tokenomics": {
        "tokenizer_bloat_factor": 1.04,
        "default_reasoning_multiplier": 1.2,
        "supports_reasoning_effort": false,
        "supports_prompt_caching": true,
        "supports_batch_api": true
      },
      "performance": {
        "ttft_ms": 195,
        "output_tps": 215,
        "coding_index": 89.7,
        "reasoning_index": 88.5,
        "agentic_tool_score": 89.3,
        "uptime_30d_pct": 99.88
      },
      "use_cases": [
        "realtime-voice-chat",
        "high-volume-rag-extraction",
        "eu-sovereign-self-hosted"
      ],
      "affiliate_partner_slug": "fireworks-fast-inference",
      "summary": "Blazingly fast 215 tok/s open-weight MoE model with 1M context window — ideal for voice agents and high-concurrency classification pipelines.",
      "derived_finops": {
        "cache_break_even_ratio": 0,
        "benchmark_10k_agent_tasks_usd": 238.76
      }
    },
    {
      "model_key": "meta/llama-4-scout-109b",
      "slug": "llama-4-scout-109b",
      "name": "Llama 4 Scout 109B",
      "provider": "Meta (Hosted Cloud)",
      "provider_slug": "meta",
      "tier": "Open-Weight Efficiency",
      "license": "Llama 4 Community",
      "release_date": "2026-05",
      "context_window": 10000000,
      "max_output_tokens": 16384,
      "pricing": {
        "input_per_1m_usd": 0.15,
        "output_per_1m_usd": 0.5,
        "cache_write_multiplier": 1,
        "cache_read_multiplier": 0.5,
        "batch_discount_multiplier": 0.5,
        "long_context_threshold_tokens": 500000,
        "long_context_input_multiplier": 1.6,
        "long_context_output_multiplier": 1.3,
        "eu_residency_markup": 1.08
      },
      "tokenomics": {
        "tokenizer_bloat_factor": 1.04,
        "default_reasoning_multiplier": 1,
        "supports_reasoning_effort": false,
        "supports_prompt_caching": true,
        "supports_batch_api": true
      },
      "performance": {
        "ttft_ms": 170,
        "output_tps": 260,
        "coding_index": 83.4,
        "reasoning_index": 82.1,
        "agentic_tool_score": 84,
        "uptime_30d_pct": 99.9
      },
      "use_cases": [
        "long-context-codebase-audit",
        "high-volume-rag-extraction",
        "realtime-voice-chat"
      ],
      "affiliate_partner_slug": "fireworks-fast-inference",
      "summary": "Ultra-long 10M context window specialist built for repository-wide indexing, log forensics, and multi-document legal/financial sweeps at $0.15/1M input.",
      "derived_finops": {
        "cache_break_even_ratio": 0,
        "benchmark_10k_agent_tasks_usd": 129.25
      }
    },
    {
      "model_key": "mistral/mistral-large-3",
      "slug": "mistral-large-3",
      "name": "Mistral Large 3",
      "provider": "Mistral AI",
      "provider_slug": "mistral",
      "tier": "Balanced Production",
      "license": "Proprietary / EU Sovereign",
      "release_date": "2026-07",
      "context_window": 256000,
      "max_output_tokens": 32768,
      "pricing": {
        "input_per_1m_usd": 1.8,
        "output_per_1m_usd": 5.4,
        "cache_write_multiplier": 1,
        "cache_read_multiplier": 0.3,
        "batch_discount_multiplier": 0.5,
        "long_context_threshold_tokens": 128000,
        "long_context_input_multiplier": 1.5,
        "long_context_output_multiplier": 1.25,
        "eu_residency_markup": 1
      },
      "tokenomics": {
        "tokenizer_bloat_factor": 1.07,
        "default_reasoning_multiplier": 1.4,
        "supports_reasoning_effort": true,
        "supports_prompt_caching": true,
        "supports_batch_api": true
      },
      "performance": {
        "ttft_ms": 310,
        "output_tps": 124,
        "coding_index": 91.9,
        "reasoning_index": 91,
        "agentic_tool_score": 92.4,
        "uptime_30d_pct": 99.92
      },
      "use_cases": [
        "eu-sovereign-self-hosted",
        "autonomous-coding-agents",
        "high-volume-rag-extraction"
      ],
      "affiliate_partner_slug": "scaleway-eu-ai",
      "summary": "Native Paris/EU data-residency flagship with 0% GDPR sovereignty surcharge (1.00x EU markup) and strong multilingual European compliance performance.",
      "derived_finops": {
        "cache_break_even_ratio": 0,
        "benchmark_10k_agent_tasks_usd": 1430.83
      }
    },
    {
      "model_key": "mistral/codestral-26-05",
      "slug": "codestral-26-05",
      "name": "Codestral 26.05",
      "provider": "Mistral AI",
      "provider_slug": "mistral",
      "tier": "Fast Frontier Runtime",
      "license": "Mistral Non-Production / Commercial API",
      "release_date": "2026-05",
      "context_window": 256000,
      "max_output_tokens": 16384,
      "pricing": {
        "input_per_1m_usd": 0.3,
        "output_per_1m_usd": 0.9,
        "cache_write_multiplier": 1,
        "cache_read_multiplier": 0.3,
        "batch_discount_multiplier": 0.5,
        "long_context_threshold_tokens": 128000,
        "long_context_input_multiplier": 1.5,
        "long_context_output_multiplier": 1.25,
        "eu_residency_markup": 1
      },
      "tokenomics": {
        "tokenizer_bloat_factor": 1.03,
        "default_reasoning_multiplier": 1,
        "supports_reasoning_effort": false,
        "supports_prompt_caching": true,
        "supports_batch_api": true
      },
      "performance": {
        "ttft_ms": 145,
        "output_tps": 240,
        "coding_index": 91.2,
        "reasoning_index": 83.5,
        "agentic_tool_score": 87.8,
        "uptime_30d_pct": 99.94
      },
      "use_cases": [
        "autonomous-coding-agents",
        "eu-sovereign-self-hosted"
      ],
      "affiliate_partner_slug": "scaleway-eu-ai",
      "summary": "Sub-150ms Fill-in-the-Middle (FIM) IDE autocomplete and diff-generation specialist with native EU hosting and zero residency markup.",
      "derived_finops": {
        "cache_break_even_ratio": 0,
        "benchmark_10k_agent_tasks_usd": 212.87
      }
    },
    {
      "model_key": "qwen/qwen-3-5-235b",
      "slug": "qwen-3-5-235b",
      "name": "Qwen 3.5 235B",
      "provider": "Alibaba Qwen (Cloud Hosted)",
      "provider_slug": "qwen",
      "tier": "Open-Weight Efficiency",
      "license": "Apache 2.0",
      "release_date": "2026-08",
      "context_window": 262144,
      "max_output_tokens": 32768,
      "pricing": {
        "input_per_1m_usd": 0.22,
        "output_per_1m_usd": 0.75,
        "cache_write_multiplier": 1,
        "cache_read_multiplier": 0.25,
        "batch_discount_multiplier": 0.5,
        "long_context_threshold_tokens": 131072,
        "long_context_input_multiplier": 1.6,
        "long_context_output_multiplier": 1.4,
        "eu_residency_markup": 1.15
      },
      "tokenomics": {
        "tokenizer_bloat_factor": 1.03,
        "default_reasoning_multiplier": 1.65,
        "supports_reasoning_effort": true,
        "supports_prompt_caching": true,
        "supports_batch_api": true
      },
      "performance": {
        "ttft_ms": 280,
        "output_tps": 142,
        "coding_index": 93.1,
        "reasoning_index": 92.8,
        "agentic_tool_score": 91.9,
        "uptime_30d_pct": 99.82
      },
      "use_cases": [
        "autonomous-coding-agents",
        "high-volume-rag-extraction",
        "eu-sovereign-self-hosted"
      ],
      "affiliate_partner_slug": "together-gpu-cloud",
      "summary": "Apache-2.0 licensed hybrid thinking MoE model delivering 93+ coding performance at under $0.80/1M output tokens on dedicated inference clouds.",
      "derived_finops": {
        "cache_break_even_ratio": 0,
        "benchmark_10k_agent_tasks_usd": 175.81
      }
    }
  ],
  "integrity_sha256": "08871b1d8e752608821acd9bf9686c676ab9c9174c09e5d2ac6f48d0412ee279"
}
