Model Changelog
New models, price moves, and retirements — generated from the same registry that runs every benchmark. If it changed here, it changed for your tasks too.
Why this matters: model choice isn't a one-time decision. Prices move, models get superseded, and quality shifts between versions. When something on this page touches a model you rely on, that's your cue to re-run your benchmark.
2026-08-20
📈 Price: DeepSeek-V4-Flash input $0.14 → $0.44 per 1M (+214.3%) — more expensive
📈 Price: DeepSeek-V4-Flash output $0.28 → $1.32 per 1M (+371.4%) — more expensive
📈 Price: DeepSeek-V4-Flash cached input $0.0028 → $0.014 per 1M (+400.0%) — more expensive
📈 Price: DeepSeek-V4-Pro input $0.435 → $1.32 per 1M (+203.4%) — more expensive
📈 Price: DeepSeek-V4-Pro output $0.87 → $3.96 per 1M (+355.2%) — more expensive
📈 Price: DeepSeek-V4-Pro cached input $0.003625 → $0.044 per 1M (+1113.8%) — more expensive
📈 Price: DeepSeek-V4-Flash (Non-Reasoning) input $0.14 → $0.44 per 1M (+214.3%) — more expensive
📈 Price: DeepSeek-V4-Flash (Non-Reasoning) output $0.28 → $1.32 per 1M (+371.4%) — more expensive
📈 Price: DeepSeek-V4-Flash (Non-Reasoning) cached input $0.0028 → $0.014 per 1M (+400.0%) — more expensive
📉 Price: GPT-5.6 Terra input $2.5 → $2 per 1M (-20.0%) — cheaper
📉 Price: GPT-5.6 Terra output $15 → $12 per 1M (-20.0%) — cheaper
📉 Price: GPT-5.6 Terra cached input $0.25 → $0.2 per 1M (-20.0%) — cheaper
📉 Price: GPT-5.6 Luna input $1 → $0.2 per 1M (-80.0%) — cheaper
📉 Price: GPT-5.6 Luna output $6 → $1.2 per 1M (-80.0%) — cheaper
📉 Price: GPT-5.6 Luna cached input $0.1 → $0.02 per 1M (-80.0%) — cheaper
🚫 Retired/suspended: Qwen 2.5 7B Instruct Turbo API (Qwen (Alibaba))
🚫 Retired/suspended: GPT-5.2 Chat (OpenAI)
🚫 Retired/suspended: GPT-5.3 Chat (OpenAI)
🚫 Retired/suspended: Kimi K2.6 Fp4 (Moonshot AI)
🚫 Retired/suspended: Kimi K2.7 Code (Moonshot AI)
2026-07-24
➕ New: Claude Opus 5 (Anthropic) — $5/$25 per 1M, 1M context
➕ New: Grok 4.5 (xAI) — $2/$6 per 1M, 500K context
📉 Price: DeepSeek-V4-Flash cached input $0.028 → $0.0028 per 1M (-90.0%) — cheaper
📉 Price: DeepSeek-V4-Flash (Non-Reasoning) cached input $0.028 → $0.0028 per 1M (-90.0%) — cheaper
🚫 Retired/suspended: GPT-5 Chat (OpenAI)
🚫 Retired/suspended: Llama 3 8B Instruct Lite API (Meta)
🚫 Retired/suspended: Qwen3-235B A22B Instruct (tput) (Qwen (Alibaba))
🚫 Retired/suspended: GPT-5-Codex (OpenAI)
🚫 Retired/suspended: GPT-5.1 Chat (OpenAI)
🚫 Retired/suspended: GPT-5.1-Codex (OpenAI)
🚫 Retired/suspended: GPT-5.1 Codex Mini (OpenAI)
🚫 Retired/suspended: GPT-5.1 Codex Max (OpenAI)
🚫 Retired/suspended: GPT-5.2-Codex (OpenAI)
🚫 Retired/suspended: GLM-5.1 (Zhipu AI)
2026-07-09
➕ New: GPT-5.6 Sol (OpenAI) — $5/$30 per 1M, 1.05M context
➕ New: GPT-5.6 Terra (OpenAI) — $2.5/$15 per 1M, 1.05M context
➕ New: GPT-5.6 Luna (OpenAI) — $1/$6 per 1M, 1.05M context
2026-07-02
➕ New: Claude Fable 5 (Anthropic) — $10/$50 per 1M, 1M context
🗑️ Removed from registry: ⚠️ Claude Fable 5 (Anthropic)
2026-06-30
➕ New: Claude Sonnet 5 (Anthropic) — $3/$15 per 1M, 1M context
2026-06-24
➕ New: MiniMax-M3 (MiniMax) — $0.3/$1.2 per 1M, 1M context
➕ New: GLM-5.2 (Zhipu AI) — $1.4/$4.4 per 1M, 262.144K context
➕ New: Qwen3.7 Max (Qwen (Alibaba)) — $1.25/$3.75 per 1M, 1M context
➕ New: Kimi K2.7 Code (Moonshot AI) — $0.95/$4 per 1M, 262.144K context
2026-06-16
🚫 Retired/suspended: Qwen3.5 397B A17b (Qwen (Alibaba))
2026-06-13
➕ New: ⚠️ Claude Fable 5 (Anthropic) — $10/$50 per 1M, 1M context
🗑️ Removed from registry: Claude Fable 5 (Anthropic)
2026-06-12
📉 Price: DeepSeek-V4-Pro input $1.74 → $0.435 per 1M (-75.0%) — cheaper
📉 Price: DeepSeek-V4-Pro output $3.48 → $0.87 per 1M (-75.0%) — cheaper
📉 Price: DeepSeek-V4-Pro cached input $0.145 → $0.003625 per 1M (-97.5%) — cheaper
2026-06-09
➕ New: Claude Fable 5 (Anthropic) — $10/$50 per 1M, 1M context
➕ New: Claude Mythos 5 (Anthropic) — $10/$50 per 1M, 1M context not yet enabled
2026-05-29
➕ New: Mistral Medium 3.5 (Mistral) — $1.5/$7.5 per 1M, 128K context
🚫 Retired/suspended: Devstral 2.0 (Mistral)
🚫 Retired/suspended: Magistral Medium 1.1 (Mistral)
🚫 Retired/suspended: Mistral Nemo 12B (Mistral)
🗑️ Removed from registry: Mistral Medium 3.1 (Mistral)
2026-05-28
➕ New: Claude Opus 4.8 (Anthropic) — $5/$25 per 1M, 200K context
2026-05-26
📈 Price: Llama 3 8B Instruct Lite API input $0.1 → $0.14 per 1M (+40.0%) — more expensive
📈 Price: Llama 3 8B Instruct Lite API output $0.1 → $0.14 per 1M (+40.0%) — more expensive
📈 Price: Llama 3.3 70B Instruct Turbo API input $0.88 → $1.04 per 1M (+18.2%) — more expensive
📈 Price: Llama 3.3 70B Instruct Turbo API output $0.88 → $1.04 per 1M (+18.2%) — more expensive
📈 Price: Qwen3.5 9B FP8 input $0.1 → $0.17 per 1M (+70.0%) — more expensive
📈 Price: Qwen3.5 9B FP8 output $0.15 → $0.25 per 1M (+66.7%) — more expensive
2026-05-24
🚫 Retired/suspended: Qwen3-Coder-480B A35B Instruct (Qwen (Alibaba))
2026-05-19
➕ New: Gemini 3.5 Flash (Google (Gemini)) — $1.5/$9 per 1M, 1.04858M context
2026-05-07
➕ New: GLM-5.1 (Zhipu AI) — $1.4/$4.4 per 1M, 202.8K context
➕ New: Kimi K2.6 Fp4 (Moonshot AI) — $1.2/$4.5 per 1M, 262.144K context
🚫 Retired/suspended: Kimi K2.5 (Moonshot AI)
🗑️ Removed from registry: GLM-5.1 (Zhipu AI)
2026-05-01
➕ New: Grok 4.2 (Non-Reasoning) (xAI) — $1.25/$2.5 per 1M, 2M context
➕ New: Grok 4.2 (Reasoning) (xAI) — $1.25/$2.5 per 1M, 2M context
➕ New: Grok 4.3 (xAI) — $1.25/$2.5 per 1M, 2M context
2026-04-24
➕ New: DeepSeek-V4-Flash (DeepSeek) — $0.14/$0.28 per 1M, 1M context
➕ New: DeepSeek-V4-Pro (DeepSeek) — $1.74/$3.48 per 1M, 1M context
➕ New: DeepSeek-V4-Flash (Non-Reasoning) (DeepSeek) — $0.14/$0.28 per 1M, 1M context
➕ New: GPT-5.5 (OpenAI) — $5/$30 per 1M, 1.05M context
➕ New: GPT-5.5 Pro (OpenAI) — $30/$180 per 1M, 1.05M context
🚫 Retired/suspended: DeepSeek-V3.2-Exp (Non-thinking Mode) (DeepSeek)
🚫 Retired/suspended: DeepSeek-V3.2-Exp (Thinking Mode) (DeepSeek)
2026-04-21
➕ New: Cogito v2.1 671B (Deep Cogito) — $1.25/$1.25 per 1M, 163.8K context
🚫 Retired/suspended: Llama 3.1 405B Instruct Turbo API (Meta)
🚫 Retired/suspended: Llama 4 Scout Instruct API (Meta)
🚫 Retired/suspended: Cogito v2 preview 405B (Deep Cogito)
🚫 Retired/suspended: Cogito v2 preview 70B (Deep Cogito)
2026-04-16
➕ New: GLM-5.1 (Zhipu AI) — $1.4/$4.4 per 1M, 202.8K context
➕ New: Claude Opus 4.7 (Anthropic) — $5/$25 per 1M, 200K context
🗑️ Removed from registry: GLM-5.1 (Zhipu AI)
2026-04-09
➕ New: GLM-5.1 (Zhipu AI) — $1.4/$4.4 per 1M, 202.8K context
🗑️ Removed from registry: GLM-5-FP4 (Zhipu AI)
2026-03-26
➕ New: MiniMax-M2.7 (MiniMax) — $0.3/$1.2 per 1M, 196.608K context
➕ New: MiniMax-M2.7-highspeed (MiniMax) — $0.6/$2.4 per 1M, 196.608K context
📉 Price: MiniMax-M2.5-lightning output $4.8 → $2.4 per 1M (-50.0%) — cheaper
2026-03-25
🚫 Retired/suspended: Llama 4 Maverick Instruct API (Meta)
2026-03-20
🚫 Retired/suspended: GLM-4.5-Air (Zhipu AI)
🚫 Retired/suspended: GLM 4.7 Fp8 (Zhipu AI)
🚫 Retired/suspended: Qwen3 Next 80B A3b Instruct (Qwen (Alibaba))
2026-03-17
➕ New: GPT-5.4 mini (OpenAI) — $0.75/$4.5 per 1M, 400K context
➕ New: GPT-5.4 nano (OpenAI) — $0.2/$1.25 per 1M, 400K context
📈 Price: Mistral Small 4 input $0.1 → $0.15 per 1M (+50.0%) — more expensive
📈 Price: Mistral Small 4 output $0.3 → $0.6 per 1M (+100.0%) — more expensive
🪟 Context window: Mistral Small 4 128K → 256K tokens
2026-03-13
➕ New: Qwen3.5 397B A17b (Qwen (Alibaba)) — $0.6/$3.6 per 1M, 262.144K context
➕ New: Qwen3.5 9B FP8 (Qwen (Alibaba)) — $0.1/$0.15 per 1M, 262.144K context
🚫 Retired/suspended: Llama 3.1 8B Instruct Turbo API (Meta)
🗑️ Removed from registry: Qwen3.5 397B A17b (Qwen (Alibaba))
2026-03-09
🚫 Retired/suspended: Gemini 3 Pro (Google (Gemini))
2026-03-06
➕ New: Kimi K2.5 (Moonshot AI) — $0.5/$2.8 per 1M, 262.144K context
🚫 Retired/suspended: Llama 3.2 3B Instruct Turbo API (Meta)
🚫 Retired/suspended: Qwen3 235B A22B Thinking 2507 (FP8) API (Qwen (Alibaba))
🚫 Retired/suspended: Marin 8B Instruct (Marin)
🚫 Retired/suspended: Kimi K2 (Moonshot AI)
🚫 Retired/suspended: Kimi K2 Thinking (Moonshot AI)
🚫 Retired/suspended: Qwen3 Next 80B A3b Thinking (Qwen (Alibaba))
2026-03-05
➕ New: GPT-5.4 (OpenAI) — $2.5/$15 per 1M, 1.05M context
➕ New: GPT-5.4 Pro (OpenAI) — $30/$180 per 1M, 1.05M context
2026-03-03
➕ New: Devstral 2.0 (Mistral) — $0.4/$2 per 1M, 128K context
➕ New: GPT-5.3 Chat (OpenAI) — $1.75/$14 per 1M, 128K context
➕ New: Gemini 3.1 Flash-Lite (Google (Gemini)) — $0.25/$1.5 per 1M, 1.04858M context
🗑️ Removed from registry: Devstral 2.1 (Mistral)
2026-02-22
🚫 Retired/suspended: Claude Haiku 3.5 (Anthropic)
2026-02-21
2026-02-19
➕ New: Gemini 3.1 Pro (Google (Gemini)) — $2/$12 per 1M, 1.04858M context
2026-02-17
➕ New: Qwen3.5 397B A17b (Qwen (Alibaba)) — $0.6/$3.6 per 1M, 262.144K context
➕ New: Claude Sonnet 4.6 (Anthropic) — $3/$15 per 1M, 200K context
2026-02-14
➕ New: GLM-5-FP4 (Zhipu AI) — $1/$3.2 per 1M, 202.752K context
2026-02-13
➕ New: MiniMax-M2.5 (MiniMax) — $0.3/$1.2 per 1M, 196.608K context
➕ New: MiniMax-M2.5-lightning (MiniMax) — $0.6/$4.8 per 1M, 196.608K context
📈 Price: MiniMax-M2.1-lightning input $0.3 → $0.6 per 1M (+100.0%) — more expensive
📈 Price: MiniMax-M2.1-lightning output $2.4 → $4.8 per 1M (+100.0%) — more expensive
📈 Price: MiniMax-M2.1-lightning cached input $0.03 → $0.06 per 1M (+100.0%) — more expensive
2026-02-12
🚫 Retired/suspended: Trinity Mini (Arcee AI)
2026-02-05
➕ New: Claude Opus 4.6 (Anthropic) — $5/$25 per 1M, 200K context
2026-02-03
✅ Back online: Command A (Cohere)
✅ Back online: Command R (Cohere)
✅ Back online: Command R7B (Cohere)
2026-01-31
📈 Price: DeepSeek-V3.2-Exp (Non-thinking Mode) input $0.028 → $0.28 per 1M (+900.0%) — more expensive
📈 Price: DeepSeek-V3.2-Exp (Thinking Mode) input $0.028 → $0.28 per 1M (+900.0%) — more expensive
📉 Price: Gemini 2.5 Flash cached input $0.075 → $0.03 per 1M (-60.0%) — cheaper
📉 Price: Gemini 2.5 Flash-Lite cached input $0.025 → $0.01 per 1M (-60.0%) — cheaper
📉 Price: Gemini 2.5 Pro cached input $0.31 → $0.125 per 1M (-59.7%) — cheaper
📈 Price: Ministral 3B input $0.04 → $0.1 per 1M (+150.0%) — more expensive
📈 Price: Ministral 3B output $0.04 → $0.1 per 1M (+150.0%) — more expensive
📈 Price: Ministral 8B input $0.1 → $0.15 per 1M (+50.0%) — more expensive
📈 Price: Ministral 8B output $0.1 → $0.15 per 1M (+50.0%) — more expensive
📉 Price: Mistral Large 3 input $2 → $0.5 per 1M (-75.0%) — cheaper
📉 Price: Mistral Large 3 output $6 → $1.5 per 1M (-75.0%) — cheaper
🪟 Context window: Codestral 2508 256K → 128K tokens
🪟 Context window: Devstral Medium 131.072K → 128K tokens
🪟 Context window: Magistral Medium 1.1 40.96K → 128K tokens
🪟 Context window: Magistral Small 1.1 40.96K → 128K tokens
🪟 Context window: Ministral 3B 32.768K → 256K tokens
🪟 Context window: Ministral 8B 131.072K → 256K tokens
🪟 Context window: Mistral Large 3 131.072K → 256K tokens
🪟 Context window: Mistral Medium 3.1 131.072K → 128K tokens
🪟 Context window: Mistral Small 3.2 131.072K → 128K tokens
2026-01-27
➕ New: Gemini Robotics-ER 1.5 Preview (Google (Gemini)) — $0.3/$2.5 per 1M, 1.04858M context
➕ New: Gemini 2.5 Computer Use Preview (Google (Gemini)) — $1.25/$10 per 1M, 128K context not yet enabled
2026-01-25
➕ New: Claude Sonnet 4.5 (Anthropic) — $3/$15 per 1M, 200K context
➕ New: Claude Haiku 4.5 (Anthropic) — $1/$5 per 1M, 200K context
➕ New: Claude Opus 4.5 (Anthropic) — $5/$25 per 1M, 200K context
➕ New: Gemini 3 Flash (Google (Gemini)) — $0.5/$3 per 1M, 1.04858M context
➕ New: Gemini 3 Pro (Google (Gemini)) — $2/$12 per 1M, 1.04858M context
➕ New: GPT-5-Codex (OpenAI) — $1.25/$10 per 1M, 400K context
➕ New: GPT-5 pro (OpenAI) — $15/$120 per 1M, 400K context
➕ New: GPT-5.1 (OpenAI) — $1.25/$10 per 1M, 400K context
➕ New: GPT-5.1 Chat (OpenAI) — $1.25/$10 per 1M, 128K context
➕ New: GPT-5.1-Codex (OpenAI) — $1.25/$10 per 1M, 400K context
➕ New: GPT-5.1 Codex Mini (OpenAI) — $0.25/$2 per 1M, 400K context
➕ New: GPT-5.1 Codex Max (OpenAI) — $1.25/$10 per 1M, 400K context
➕ New: GPT-5.2 (OpenAI) — $1.75/$14 per 1M, 400K context
➕ New: GPT-5.2 Chat (OpenAI) — $1.75/$14 per 1M, 128K context
➕ New: GPT-5.2 pro (OpenAI) — $21/$168 per 1M, 400K context
➕ New: GPT-5.2-Codex (OpenAI) — $1.75/$14 per 1M, 400K context
➕ New: Ministral 14B (Mistral) — $0.2/$0.2 per 1M, 256K context
➕ New: Mistral Nemo 12B (Mistral) — $0.15/$0.15 per 1M, 128K context
➕ New: Grok 4 Fast (Non-Reasoning) (xAI) — $0.2/$0.5 per 1M, 2M context
➕ New: Grok 4 Fast (xAI) — $0.2/$0.5 per 1M, 2M context
➕ New: Grok 4.1 Fast (Non-Reasoning) (xAI) — $0.2/$0.5 per 1M, 2M context
➕ New: Grok 4.1 Fast (xAI) — $0.2/$0.5 per 1M, 2M context
➕ New: Grok Code Fast 1 (xAI) — $0.2/$1.5 per 1M, 256K context
➕ New: Glm 4.6 Fp8 (Zhipu AI) — $0.6/$2.2 per 1M, 202.752K context
➕ New: GLM 4.7 Fp8 (Zhipu AI) — $0.45/$2 per 1M, 202.752K context
➕ New: Kimi K2 Thinking (Moonshot AI) — $1.2/$4 per 1M, 262.144K context
➕ New: MiniMax-M2 (MiniMax) — $0.3/$1.2 per 1M, 196.608K context
➕ New: MiniMax-M2-her (MiniMax) — $0.3/$1.2 per 1M, 32.768K context
➕ New: MiniMax-M2.1 (MiniMax) — $0.3/$1.2 per 1M, 196.608K context
➕ New: MiniMax-M2.1-lightning (MiniMax) — $0.3/$2.4 per 1M, 196.608K context
➕ New: Trinity Mini (Arcee AI) — $0.05/$0.15 per 1M, 131.072K context
➕ New: Qwen3 Next 80B A3b Thinking (Qwen (Alibaba)) — $0.15/$1.5 per 1M, 262.144K context
➕ New: Qwen3 Next 80B A3b Instruct (Qwen (Alibaba)) — $0.15/$1.5 per 1M, 262.144K context
🪟 Context window: Kimi K2-Instruct 0905 131.072K → 262.144K tokens
🚫 Retired/suspended: OpenAI o1-mini (OpenAI)
🚫 Retired/suspended: Claude Sonnet 3.5 (Anthropic)
🚫 Retired/suspended: Gemini 2.0 Flash (Google (Gemini))
🚫 Retired/suspended: Gemini 2.0 Flash-Lite (Google (Gemini))
🚫 Retired/suspended: Command A (Cohere)
🚫 Retired/suspended: Command R (Cohere)
🚫 Retired/suspended: Command R7B (Cohere)
🚫 Retired/suspended: Sonar Reasoning (Perplexity)
🚫 Retired/suspended: Grok 2 (xAI)
🚫 Retired/suspended: Grok 3 Fast (xAI)
🚫 Retired/suspended: Grok 3 Mini Fast (xAI)
🚫 Retired/suspended: Llama 3 70B Instruct Reference API (Meta)
🚫 Retired/suspended: Qwen 2.5 Coder 32B Instruct API (Qwen (Alibaba))
🚫 Retired/suspended: QwQ-32B (Qwen (Alibaba))
🚫 Retired/suspended: Arcee AI Coder-Large (Arcee AI)
🚫 Retired/suspended: Arcee AI Maestro Reasoning (Arcee AI)
🚫 Retired/suspended: Arcee AI Spotlight (Arcee AI)
🚫 Retired/suspended: Arcee AI Virtuoso-Large (Arcee AI)
🚫 Retired/suspended: Cogito v2 preview 671B MoE (Deep Cogito)
🚫 Retired/suspended: gpt-oss-20B (OpenAI)
🗑️ Removed from registry: Devstral Small 1.1 (Mistral)
Did One of These Changes Touch Your Model?
Re-run your benchmark and see if a cheaper or newer model now wins.
Real API calls, ranked results in minutes. Free tier available.