AI 模型變更日誌 · 2026-10-21

2026-10-21 當日紀錄 500 個事件。

能力欄位更新 (314)

aihubmix Capabilities updated: LongCat-2.0
aimlapi Capabilities updated: Aion-RP 1.0 (8B)
aimlapi Capabilities updated: Schematron V2 Small
aimlapi Capabilities updated: Schematron V2 Turbo
Alibaba (Qwen)Capabilities updated: Qwen3.8 27B
alibaba-cn Capabilities updated: QVQ Max
alibaba-cn Capabilities updated: DeepSeek R1 Distill Qwen 1.5B
Alibaba (Qwen)Capabilities updated: Qwen Flash
Alibaba (Qwen)Capabilities updated: Qwen Long
Alibaba (Qwen)Capabilities updated: Qwen Max
Alibaba (Qwen)Capabilities updated: Qwen-Omni Turbo
Alibaba (Qwen)Capabilities updated: Qwen-Omni Turbo Realtime
Alibaba (Qwen)Capabilities updated: Qwen Plus
Alibaba (Qwen)Capabilities updated: Qwen Plus Character
Alibaba (Qwen)Capabilities updated: Qwen Plus Character (Japanese)
Alibaba (Qwen)Capabilities updated: QVQ Max
Alibaba (Qwen)Capabilities updated: Qwen2.5-Omni 7B
Alibaba (Qwen)Capabilities updated: Qwen2.5-VL 72B Instruct
Alibaba (Qwen)Capabilities updated: Qwen2.5-VL 7B Instruct
Alibaba (Qwen)Capabilities updated: Qwen3.5 122B-A10B
Alibaba (Qwen)Capabilities updated: Qwen3.5 27B
Alibaba (Qwen)Capabilities updated: Qwen3.5 35B-A3B
Alibaba (Qwen)Capabilities updated: Qwen3.5 397B-A17B
Alibaba (Qwen)Capabilities updated: Qwen3.5 Plus
Alibaba (Qwen)Capabilities updated: Qwen3.6 27B
Alibaba (Qwen)Capabilities updated: Qwen3.6 35B
Alibaba (Qwen)Capabilities updated: Qwen3.6 35B-A3B
Alibaba (Qwen)Capabilities updated: Qwen3.6 Plus
Alibaba (Qwen)Capabilities updated: Qwen3.7 Max
Alibaba (Qwen)Capabilities updated: Qwen3.8 Flash Next
Alibaba (Qwen)Capabilities updated: Qwen3.8 Max Preview
Alibaba (Qwen)Capabilities updated: Qwen3 Coder Flash
Alibaba (Qwen)Capabilities updated: Qwen3 Embedding 4B
Alibaba (Qwen)Capabilities updated: Qwen3-LiveTranslate Flash Realtime
Alibaba (Qwen)Capabilities updated: Qwen3-Max-Preview
Alibaba (Qwen)Capabilities updated: Qwen3-Omni Flash
Alibaba (Qwen)Capabilities updated: Qwen3-Omni Flash Realtime
Alibaba (Qwen)Capabilities updated: Qwen3-VL 235B-A22B
Alibaba (Qwen)Capabilities updated: Qwen3-VL 30B-A3B
Alibaba (Qwen)Capabilities updated: Qwen3-VL Plus
Alibaba (Qwen)Capabilities updated: Qwen-VL Max
Alibaba (Qwen)Capabilities updated: Qwen-VL OCR
Alibaba (Qwen)Capabilities updated: Qwen-VL Plus
Anthropic Capabilities updated: Claude Haiku 5.5 (AU)
Anthropic Capabilities updated: Claude Opus 4.7 (AU)
Anthropic Capabilities updated: Claude Opus 4.8 (AU)
Anthropic Capabilities updated: Claude Opus 5 (AU)
Anthropic Capabilities updated: Claude Opus 5.5 (AU)
Anthropic Capabilities updated: Claude Sonnet 5 (AU)
Anthropic Capabilities updated: Anthropic: Claude 3.5 Haiku
Anthropic Capabilities updated: Claude 3.5 Sonnet
Anthropic Capabilities updated: Anthropic: Claude 3.7 Sonnet
Anthropic Capabilities updated: Anthropic: Claude 4.5 Haiku
Anthropic Capabilities updated: Anthropic: Claude Opus 4.5
Anthropic Capabilities updated: Claude Sonnet 4.5 (latest)
Anthropic Capabilities updated: Claude Sonnet 4.6
Anthropic Capabilities updated: Anthropic: Claude Opus 4
Anthropic Capabilities updated: Claude Opus 4.1
Anthropic Capabilities updated: Claude Fable 5 (EU)
Anthropic Capabilities updated: Claude Haiku 5.5 (EU)
Anthropic Capabilities updated: Claude Opus 4.7 (EU)
Anthropic Capabilities updated: Claude Opus 4.8 (EU)
Anthropic Capabilities updated: Claude Opus 5 (EU)
Anthropic Capabilities updated: Claude Opus 5.5 (EU)
Anthropic Capabilities updated: Claude Sonnet 5 (EU)
Anthropic Capabilities updated: Claude Sonnet 5.5 (EU)
Anthropic Capabilities updated: Claude Fable 5 (Global)
Anthropic Capabilities updated: Claude Fable 5.1 (Global)
Anthropic Capabilities updated: Claude Haiku 5.5 (Global)
Anthropic Capabilities updated: Claude Opus 4.7 (Global)
Anthropic Capabilities updated: Claude Opus 4.8 (Global)
Anthropic Capabilities updated: Claude Opus 5 (Global)
Anthropic Capabilities updated: Claude Opus 5.5 (Global)
Anthropic Capabilities updated: Claude Sonnet 5 (Global)
Anthropic Capabilities updated: Claude Sonnet 5.5 (Global)
Anthropic Capabilities updated: Claude Opus 5 (India)
Anthropic Capabilities updated: Claude Haiku 5.5 (JP)
Anthropic Capabilities updated: Claude Opus 4.7 (JP)
Anthropic Capabilities updated: Claude Opus 4.8 (JP)
Anthropic Capabilities updated: Claude Opus 5 (JP)
Anthropic Capabilities updated: Claude Opus 5.5 (JP)
Anthropic Capabilities updated: Claude Sonnet 5 (JP)
Anthropic Capabilities updated: Claude Fable 5 (US)
Anthropic Capabilities updated: Claude Fable 5.1 (US)
Anthropic Capabilities updated: Claude Haiku 5.5 (US)
Anthropic Capabilities updated: Claude Opus 4.7 (US)
Anthropic Capabilities updated: Claude Opus 4.8 (US)
Anthropic Capabilities updated: Claude Opus 5 (US)
Anthropic Capabilities updated: Claude Opus 5.5 (US)
Anthropic Capabilities updated: Claude Sonnet 5 (US)
Anthropic Capabilities updated: Claude Sonnet 5.5 (US)
bfl Capabilities updated: Flux 3
ByteDance (Doubao)Capabilities updated: Seed 1.6
ByteDance (Doubao)Capabilities updated: Seed 1.8
ByteDance (Doubao)Capabilities updated: Doubao-Seed-2.0-lite
ByteDance (Doubao)Capabilities updated: Seed 2.1 Turbo
Cohere Capabilities updated: Aya Expanse 32B
Cohere Capabilities updated: Aya Vision 32B
Cohere Capabilities updated: Command A
Cohere Capabilities updated: Command A Reasoning

顯示 100 / 共 314。

上下文變動 (52)

Alibaba (Qwen)Context window: Qwen3.8 27B now 262,144 (was 131,072)
Alibaba (Qwen)Context window: Qwen3.6 35B now 16,384 (was 131,056)
Anthropic Context window: Anthropic: Claude Opus 4 now 200,000 (was 192,512)
Anthropic Context window: Anthropic: Claude Sonnet 4 now 1,000,000 (was 983,040)
DeepSeek Context window: DeepSeek OCR 2 now 4,000 (was 8,192)
DeepSeek Context window: DeepSeek-R1 now 128,000 (was 163,840)
DeepSeek Context window: DeepSeek Reasoner now 1,000,000 (was 128,000)
DeepSeek Context window: DeepSeek-V3.2 now 163,840 (was 128,000)
DeepSeek Context window: DeepSeek V4 Flash 0423 now 131,072 (was 1,000,000)
DeepSeek Context window: DeepSeek V4 Flash (free) now 200,000 (was 1,000,000)
edenai Context window: Qwen3-Next 80B-A3B Instruct (Databricks) now 262,144 (was 131,072)
edenai Context window: Qwen3-Next 80B-A3B Instruct (Databricks, EU) now 262,144 (was 131,072)
edenai Context window: GPT OSS 120B (Scaleway) now 131,072 (was 128,000)
Google Context window: gemini-2.5-flash-lite-preview-09-2025 now 1,048,576 (was 1,000,000)
Google Context window: gemini-3-pro-preview now 1,048,576 (was 1,000,000)
Google Context window: Gemma 4 12B IT now 131,072 (was 262,144)
kilo Context window: OpenRouter Free Models Router now 256,000 (was 200,000)
Meta Context window: Hermes 2 Pro Llama 3 8B now 131,072 (was 8,192)
Meta Context window: Llama 3.3 70B now 128,000 (was 131,072)
Meta Context window: Llama 4 Maverick 17B Instruct now 1,048,576 (was 8,192)
Meta Context window: Llama 4 Scout 17B Instruct now 131,072 (was 8,192)
Meta Context window: DeepSeek R1 Distill Llama 70B now 128,000 (was 32,768)
Microsoft Context window: WizardLM-2 8x22B now 65,536 (was 65,535)
Mistral Context window: Mistral-7B-Instruct-v0.3 now 65,536 (was 32,768)
Mistral Context window: Devstral 2 now 200,000 (was 262,144)
Mistral Context window: Mistral Medium 3.5 now 256,000 (was 262,144)
Mistral Context window: Mistral Small 4 now 128,000 (was 256,000)
nano-gpt Context window: GLM 5.1 Thinking now 200,000 (was 202,752)
novita Context window: GLM-4.6 now 204,800 (was 0)
novita Context window: glm-4.6v now 131,072 (was 131,000)
novita Context window: glm-4.7 now 204,800 (was 205,000)
novita Context window: GLM-5 now 202,800 (was 205,000)
novita Context window: minimax-m2.1 now 204,800 (was 205,000)
novita Context window: DeepSeek-V3.2 now 163,840 (was 128,000)
NVIDIA Context window: Nemotron 3 Ultra 550B A55B now 262,144 (was 131,072)
NVIDIA Context window: Nemotron 3 Ultra 550B A55B now 1,048,576 (was 262,144)
OpenAI Context window: GPT-4o mini Transcribe now 16,000 (was 0)
OpenAI Context window: GPT-4o Transcribe now 16,000 (was 0)
OpenAI Context window: GPT-5 Chat (latest) now 128,000 (was 400,000)
openrouter Context window: Nemotron 3 Super (free) now 262,144 (was 1,000,000)
vercel Context window: Codestral (latest) now 128,000 (was 256,000)
vercel Context window: Gemini 3.5 Transcribe now 98,304 (was 0)
vercel Context window: Gemini 3.8 Flash-Lite TTS now 8,192 (was 0)
vercel Context window: Gemini 3.8 Flash TTS now 8,192 (was 0)
vercel Context window: Gemini 3.8 Live now 131,072 (was 0)
vercel Context window: Gemini 3.8 Live Extended Thinking now 131,072 (was 0)
vercel Context window: GPT-Realtime-1.5 now 32,000 (was 0)
vercel Context window: gpt-realtime-2 now 128,000 (was 0)
vercel Context window: GPT-Realtime mini now 32,000 (was 0)
vercel Context window: gpt-realtime-whisper now 16,000 (was 0)
vercel Context window: Grok TTS now 15,000 (was 0)
vercel Context window: Jev now 32,000 (was 64,000)

輸出上限變動 (83)

Alibaba (Qwen)Max output: Qwen3.8 27B updated
Alibaba (Qwen)Max output: Qwen3.5 9B updated
Alibaba (Qwen)Max output: Qwen3.6 35B updated
Alibaba (Qwen)Max output: Qwen3.8 2.4T A95B updated
Alibaba (Qwen)Max output: Qwen3.8 27B updated
Alibaba (Qwen)Max output: Qwen3.8 Flash Next updated
Alibaba (Qwen)Max output: Qwen3 Embedding 0.6B updated
Alibaba (Qwen)Max output: Qwen3 Embedding 4B updated
Alibaba (Qwen)Max output: Qwen3 Embedding 8B updated
Alibaba (Qwen)Max output: Qwen3-Max-Preview updated
Anthropic Max output: Claude 3.5 Sonnet updated
Anthropic Max output: Anthropic: Claude 4.5 Haiku updated
Anthropic Max output: Claude Sonnet 4.6 updated
Anthropic Max output: Anthropic: Claude Opus 4 updated
ByteDance (Doubao)Max output: Doubao-Seed-2.0-lite updated
ByteDance (Doubao)Max output: Seed 2.0 Lite updated
ByteDance (Doubao)Max output: Seed 2.0 Mini updated
DeepSeek Max output: DeepSeek OCR 2 updated
DeepSeek Max output: DeepSeek Pro Latest updated
DeepSeek Max output: DeepSeek-R1 updated
DeepSeek Max output: DeepSeek Reasoner updated
DeepSeek Max output: DeepSeek-V3.2 updated
DeepSeek Max output: DeepSeek V4 Flash 0423 updated
DeepSeek Max output: DeepSeek V4 Flash 0731 Fast updated
DeepSeek Max output: DeepSeek V4 Flash (free) updated
DeepSeek Max output: DeepSeek V4 Flash Latest updated
Google Max output: gemini-3-pro-preview updated
Google Max output: Gemini Pro Latest updated
Google Max output: Gemma 4 12B IT updated
llmgateway-providers Max output: GPT OSS 120B (Groq) updated
Meta Max output: Hermes 2 Pro Llama 3 8B updated
Meta Max output: Llama 3.2 11B Instruct updated
Meta Max output: DeepSeek R1 Distill Llama 70B updated
Microsoft Max output: WizardLM-2 8x22B updated
Mistral Max output: Mistral: Mistral 7B Instruct v0.1 updated
Mistral Max output: Mistral-7B-Instruct-v0.3 updated
Mistral Max output: Devstral 2 updated
Mistral Max output: Magistral Small 2506 updated
Mistral Max output: Mistral Nemo updated
Mistral Max output: Mistral Small 3.1 24B updated
Mistral Max output: Mistral Small 4 updated
Moonshot AI Max output: Kimi K2.5 updated
nano-gpt Max output: GLM 5.1 Thinking updated
nano-gpt Max output: Granite 4.1 8B updated
nano-gpt Max output: Lumen Stealth updated
nano-gpt Max output: TheDrummer Skyfall 31B v4.2 updated
novita Max output: GLM-4.6 updated
novita Max output: glm-4.6v updated
novita Max output: glm-4.7 updated
novita Max output: DeepSeek-V3.2 updated
NVIDIA Max output: Nemotron 3 Ultra 550B A55B updated
NVIDIA Max output: Nemotron 3 Ultra 550B A55B updated
OpenAI Max output: GPT-4o mini Transcribe updated
OpenAI Max output: GPT-4o Transcribe updated
OpenAI Max output: GPT-5 Chat (latest) updated
OpenAI Max output: Whisper 3 Large updated
vercel Max output: Codestral (latest) updated
openrouter Max output: Nex-N2-Pro (free) updated
requesty Max output: GLM-5.1 (EU) updated
requesty Max output: Grok 4.20 (Reasoning) updated
requesty Max output: Mistral Large 4 (EU) updated
requesty Max output: Mistral Medium 3.5 (EU) updated
requesty Max output: Qwen3.5 2B updated
requesty Max output: Seed 1.8 updated
requesty Max output: Seed 2.0 Code updated
requesty Max output: Seed 2.0 Mini updated
requesty Max output: Seed 2.0 Pro updated
vercel Max output: Codestral Embed updated
vercel Max output: Embed 5 Fast updated
vercel Max output: Embed 5 Pro updated
vercel Max output: Gemini 3.5 Transcribe updated
vercel Max output: Gemini 3.8 Flash-Lite TTS updated
vercel Max output: Gemini 3.8 Flash TTS updated
vercel Max output: Gemini 3.8 Live updated
vercel Max output: Gemini 3.8 Live Extended Thinking updated
vercel Max output: GPT-Realtime-1.5 updated
vercel Max output: gpt-realtime-2 updated
vercel Max output: GPT-Realtime mini updated
vercel Max output: gpt-realtime-whisper updated
vercel Max output: Grok TTS updated
vercel Max output: Ministral 3B (latest) updated
vercel Max output: Ministral 8B (latest) updated
vercel Max output: Cohere Rerank 3.5 updated

價格變動 (36)

Alibaba (Qwen)Price update: Qwen3.5 9B now $0.05/$0.15 (was $0.04/$0.15)
Alibaba (Qwen)Price update: Qwen3.6 35B-A3B now $0.375/$2.25 (was $0.248/$1.485)
Alibaba (Qwen)Price update: Qwen3.6 Flash now $0.25/$1.5 (was $0.1875/$1.125)
DeepSeek Price update: DeepSeek Flash Latest now $0.07/$0.5 (was $0.05/$0.024)
DeepSeek Price update: DeepSeek Pro Latest now $0.225/$8 (was $0.224/$5)
DeepSeek Price update: DeepSeek-R1 now $0.5/$2.18 (was $0.4/$1.7)
DeepSeek Price update: DeepSeek-V3.1 now $0.19/$0.71 (was $0.2/$0.7)
DeepSeek Price update: DeepSeek-V4.1-Flash now $0.04/$0.08 (was $0.0352/$0.0112)
DeepSeek Price update: DeepSeek V4 Flash Latest now $0.0168/$0.10728 (was $0.00392/$0.12)
DeepSeek Price update: DeepSeek V4 Pro 0813 now $0.442/$0.884 (was $0.35/$0.8)
digitalocean Price update: Stable Diffusion 3.5 Large now $?/$? (was $0.08/$0)
digitalocean Price update: Wan2.2-T2V-A14B now $?/$? (was $0.6/$0)
edenai Price update: GPT OSS 120B (Scaleway) now $0.05/$0.25 (was $0.2245/$0.78575)
edenai Price update: GPT OSS 20B (Greenference) now $0.02/$0.1 (was $0.009/$0.045)
edenai Price update: Ministral 3 14B (Infomaniak) now $0.33645/$0.4486 (was $0.33675/$0.449)
Google Price update: Gemini Pro Latest now $2/$12 (was $2/$10)
kilo Price update: Gemini 2.5 Pro Preview 06-05 now $0.625/$5 (was $1.25/$10)
kilo Price update: DeepSeek: DeepSeek Pro Latest now $0.152/$6 (was $0.1901/$4.2)
llmgateway Price update: MiMo-V2.5-Pro now $0.522/$1.044 (was $0.435/$0.87)
llmgateway-providers Price update: GPT OSS 120B (Groq) now $0.17/$0.55 (was $0.15/$0.75)
Meta Price update: Llama-3.1-8B-Instruct now $0.025/$0.025 (was $0.02/$0.03)
Mistral Price update: Ministral 3B (latest) now $0.1/$0.1 (was $0.04/$0.04)
Mistral Price update: Ministral 8B (latest) now $0.15/$0.15 (was $0.1/$0.1)
Mistral Price update: Voxtral Small (latest) now $0.1/$0.4 (was $0.1/$0.3)
Moonshot AI Price update: Kimi Latest now $1.2/$5.5 (was $1.65/$8.25)
nano-gpt Price update: GLM 5.2 Thinking now $0.42/$1.32 (was $1.4/$4.4)
openrouter Price update: DeepSeek Pro Latest now $0.152/$6 (was $0.1901/$4.2)
pioneer Price update: GLiNER-2.5-Decide now $0.03/$0 (was $0.15/$0.15)
pioneer Price update: GLiNER 2.5 Multi now $0.03/$0 (was $0.15/$0.15)
pioneer Price update: GLiNER2 Base now $0.03/$0 (was $0.15/$0.15)
pioneer Price update: GLiNER2 Large now $0.03/$0 (was $0.15/$0.15)
pioneer Price update: GLiNER2 Multi Large now $0.03/$0 (was $0.15/$0.15)
pioneer Price update: GLiNER2 Multi now $0.03/$0 (was $0.15/$0.15)
pioneer Price update: GLiNER2 Privacy Filter PII (Multi) now $0.03/$0 (was $0.15/$0.15)
vercel Price update: Ministral 3B (latest) now $0.1/$0.1 (was $0.04/$0.04)
vercel Price update: Ministral 8B (latest) now $0.15/$0.15 (was $0.1/$0.1)

Pricing and capabilities are refreshed daily and reconciled against each provider's official documentation. Always verify critical production decisions with the provider directly.

新增模型 (15)

能力欄位更新 (314)

上下文變動 (52)

輸出上限變動 (83)

價格變動 (36)