Claude Fable 5.2 now costs less. Improvements to our serving architecture have lowered the cost of using the model, and the new rates are live.See new pricing

CHANGES

Announcements

Model access, pricing and gateway operations — every change we make, with the date it took effect.

2026

  • In progress

    Claude Fable 5.2 now costs less.

    Improvements to our serving architecture have lowered the cost of using the model, and the new rates are live.
  • In progress

    GPT-6.2 Astra and GPT-6.2 Sol are now available.

    The new Astra costs 20% less than GPT-6 Astra, with cached input 60% cheaper; Sol keeps 6.1 Sol's price with newer capabilities. Both are 85% off official prices.
  • In progress

    Gemini 4 Carbon and Hydrogen are available.

    Use Chat Completions or Anthropic Messages with streaming and tools. See the model pages for differences and supported inputs.
  • Claude Fable 5.2 is now available.Ended · 11/19/2026

    Anthropic's Mythos-class model for demanding reasoning and long-horizon agents, with a 1M-token context and 128K output.
    This campaign has ended
  • In progress

    MiniMax H3-Image is live: 4K images in about 8 s on Highspeed, a…

    MiniMax H3-Image is live: 4K images in about 8 s on Highspeed, at 50% off official pricing. Standard costs about $0.01 per 1K image.
  • GPT Image 2.6 is live.Ended · 11/20/2026

    Flare and Sunburst generate native GIF memes and 2K/4K PNG images. Explore 8 examples, copyable prompts and API guides.
    This campaign has ended
  • In progress

    Claude Opus 5.5 Kiro is now available.

  • In progress

    Claude Sonnet 5.5 Kiro is now available.

    A separate Kiro route for Claude Sonnet 5.5, with streaming and automatic tool calls.
  • In progress

    GPT-6.1 Terra is now available.

    OpenAI positions it as a fast, efficient model for visual reasoning, precise computer use and clear, natural writing. On RouteMux: $0.15 in / $0.75 out per 1M tokens (official $1 / $5).
  • In progress

    A wave of new models, and price cuts

    Up to 85% off official prices: new models and price cuts. Grok 4.6 / 4.7 are now 85% off, DeepSeek V4.1 Flash is half price and MiniMax H3 costs less; new are Gemini 4 Argon, Qwen3.8, Nano Banana 2.1 / Pro, Grok Imagine image and video and more.
  • In progress

    Grok Imagine Video 1.5 is now available.

    xAI's latest video model: text-to-video and first-frame image-to-video with sound, 1–15 sec, up to native 1080p.
  • In progress

    DeepSeek V4.1 Flash is now on the 1017 release.

    DeepSeek reports higher benchmark scores, faster task completion and shorter outputs. Same model id and pricing, no code changes needed.
  • In progress

    Pricing update from Oct 18, 04:06 UTC: with upstream GPT capacit…

    Pricing update from Oct 18, 04:06 UTC: with upstream GPT capacity reduced and costs up, GPT text models other than Luna move from 90% to 85% off list price (GPT-6.1 Sol: $0.30 in / $1.50 out per 1M tokens). Luna and image models stay unchanged.
  • In progress

    The MiniMax daily free allowance promotion has ended.

    From October 10, 2026, we no longer grant a daily free allowance for MiniMax M3, M2.7 and M2.7 Highspeed. All three models remain available and are billed from your balance at the prices on their model pages. Thanks for taking part.
  • In progress

    Claude Haiku 5.5 is now available: Anthropic's newest Haiku, rel…

    Claude Haiku 5.5 is now available: Anthropic's newest Haiku, released October 7 and the fastest model in its current lineup, built for high-volume classification, extraction and routing, with a 1M-token context and image input. $0.03 / $0.15 per 1M input / output tokens (prompts ≤100K tokens).
  • In progress

    DeepSeek V4.1 Flash is now available: DeepSeek's newest Flash mo…

    DeepSeek V4.1 Flash is now available: DeepSeek's newest Flash model, with native image input and a 1M-token context, at $0.15 / $0.60 per 1M input / output tokens. GLM-5.3 Flash now accepts image input too.
  • Pricing update from Oct 6: upstream GPT capacity has been reduce…Ended · 10/06/2026

    Pricing update from Oct 6: upstream GPT capacity has been reduced, so GPT-6 Luna and GPT-5.6 Luna go from 70% to 50% off list price (GPT-6 Luna: $0.05 in / $0.25 out per 1M tokens); other GPT models stay at 90% off. Claude Fable 5 and 5.1 have been at 70% off since Oct 2.
    This campaign has ended
  • In progress

    MiniMax H3 is now available.

    MiniMax's new open multimodal video model: text-to-video, first/last-frame image-to-video and reference images, 4–15 sec, up to 2K. Official price: $0.08/sec at 768P, $0.13/sec at 2K.
  • October Sale: GPT at half price (95% off list), Claude 80% off (…Ended · 10/06/2026

    October Sale: GPT at half price (95% off list), Claude 80% off (except Fable), +20% credit on every top-up
    This campaign has ended
  • GPT-6.1 Sol is now available.Ended · 10/18/2026

    OpenAI says it nearly matches GPT-6 Astra on agentic coding, computer use and professional work at a fifth of Astra's price. On RouteMux: $0.20 in / $1.00 out per 1M (official $2 / $10).
    This campaign has ended
  • In progress

    Claude Sonnet 5.5 is now available.

    Anthropic's newest Sonnet, released September 28, succeeds Sonnet 5 as its best balance of speed and intelligence, with adaptive thinking, a 1M-token context and image input.
  • In progress

    MiniMax M3.1 Flash Preview is now available.

    MiniMax's first M3.1 model, released September 27, is built for everyday coding and agent work: 1M-token context, image input, tool calling, and always-on thinking with five effort levels.
  • In progress

    GPT-6 Sol is now available.

    The middle tier of the GPT-6 family, released with Luna on September 22, brings most of Astra's capability to a faster model for complex coding and agents. 1.05M context.
  • In progress

    GPT-6 Luna is now available.

    The lightest GPT-6 model, released with Sol on September 22, for classification, extraction and high-volume agent steps. 1.05M context, reasoning effort from none to max.
  • In progress

    Claude Opus 5.5 is now available.

    Anthropic's new flagship, released September 22, succeeds Opus 5 for long-running agentic coding and knowledge work, with adaptive thinking, a 1M-token context and image input.
  • In progress

    Kimi K3 and Kimi K2.7 Code are now available.

    K3 is Moonshot's flagship for agentic coding and long tool chains (1M context); K2.7 Code is the coding-tuned K2 model, lighter on thinking tokens (256K context).
  • In progress

    Grok 4.7 is now available.

    xAI's new flagship, released September 21, succeeds Grok 4.6 with strong reasoning, agentic coding and tool calling, image input and a 500K-token context, over OpenAI Chat or Responses.
  • In progress

    Speech is live: text-to-speech and transcription.

    MiniMax speech-2.8-turbo (low latency) and speech-2.8-hd (high fidelity) turn text into audio; asr-1.0 transcribes audio with speaker labels and per-segment timestamps.
  • In progress

    GPT Image 2.5 is now available — OpenAI's newest image model, re…

    GPT Image 2.5 is now available — OpenAI's newest image model, released 8 September 2026. Billed per image at a flat $0.03, so prompt length and scene complexity never change the price. Each call returns one image of roughly 1.5 megapixels; the size parameter picks the aspect ratio. Requests to gpt-image-2 are now served by 2.5 at the same price, with no code changes.
  • In progress

    Gemini 3.5 Flash-Lite is now available.

    Google's lightweight model for high-volume extraction, classification and routing; thinking defaults to minimal for throughput. 1M-token context and image input.