AI Rankings

Ranking changelog

Products we add or remove, tier changes and larger rank moves, newest first. Smaller reshuffles are summarized as a count.

We updated how scores are calculated, so 147 entries changed rank or tier. How scores work

  • AddedClaude Fable 5.1now #7 among all models
  • AddedGLM-5.3-Flashnow #58 among all models
  • RemovedcamelAIData-analyst product moved to Legacy API.
  • RemovedRow64Now a real-time operations dashboard platform with no AI features — outside catalog scope
  • RemovedPlayHTMeta acqui-hire, platform sunset 2025-12-31, play.ht no longer resolves
  • RemovedLOVOChapter 7 bankruptcy 2026-05-27, hosting disabled 2026-08-05
  • RemovedGemini 3 Pro
    Retired by provider: Gemini 3 Pro only ever shipped as gemini-3-pro-preview (released Nov 18, 2025); Google's deprecations page lists its shutdown on March 9,… Full reason

    Retired by provider: Gemini 3 Pro only ever shipped as gemini-3-pro-preview (released Nov 18, 2025); Google's deprecations page lists its shutdown on March 9, 2026 (Vertex AI reportedly to March 23) with gemini-3.1-pro-preview as the replacement, and no GA gemini-3-pro model is listed. Calls to the model string now return errors. The officialUrl (gemini.google.com) is the consumer app, which now serves newer Gemini models, not this one.

  • RemovedClaude 3.5 Sonnet
    Retired by provider: Anthropic deprecated both Claude 3.5 Sonnet snapshots (claude-3-5-sonnet-20240620 and -20241022) on August 13, 2025 and retired them from… Full reason

    Retired by provider: Anthropic deprecated both Claude 3.5 Sonnet snapshots (claude-3-5-sonnet-20240620 and -20241022) on August 13, 2025 and retired them from the Claude API on October 28, 2025; requests now fail. The recommended replacement is claude-sonnet-4-6. Downstream projects (mcp-use, GitLab) have removed it, and the current Claude lineup is Sonnet 5 / Opus 5.x / Fable 5.x.

  • RemovedGemini 2 Pro
    Retired by provider: Gemini 2.0 Pro was only ever offered as the experimental gemini-2.0-pro-exp-02-05 (Feb 5, 2025); Google's Gemini API release notes list it… Full reason

    Retired by provider: Gemini 2.0 Pro was only ever offered as the experimental gemini-2.0-pro-exp-02-05 (Feb 5, 2025); Google's Gemini API release notes list it among models shut down on December 2, 2025. It was delisted in favor of Gemini 2.5 Pro in April 2025, and no 2.0 Pro model appears on the current deprecations/models listings. Retired by the provider; deprecation/shutdown date 2025-12-02.

  • RemovedKimi K2.5
    Retired by provider: Moonshot's own API model list states kimi-k2.5 was officially discontinued on August 31, 2026 and directs users to kimi-k3; the retirement… Full reason

    Retired by provider: Moonshot's own API model list states kimi-k2.5 was officially discontinued on August 31, 2026 and directs users to kimi-k3; the retirement was announced on Moonshot's Weibo on August 24, 2026 with API service ceasing the same day. Released January 27, 2026 as a 1T-parameter MoE with agent-swarm features, it has since been followed by K2.6 (April), K2.7 Code (June) and K3 (July 2026).

  • RemovedDeepSeek V3
    Retired by provider: DeepSeek's API change log shows V3 being replaced by V3.1 on Aug 21, 2025. From the V4 preview on April 24, 2026, the legacy… Full reason

    Retired by provider: DeepSeek's API change log shows V3 being replaced by V3.1 on Aug 21, 2025. From the V4 preview on April 24, 2026, the legacy deepseek-chat/deepseek-reasoner names routed to deepseek-v4-flash, and DeepSeek fully retired them on July 24, 2026 (15:59 UTC). DeepSeek's own API and app no longer serve V3; its open weights exist but the provider has sunset it.

  • RemovedKimi K2 Thinking
    Retired by provider: Moonshot's Kimi API model list states that the kimi-k2 series models, including kimi-k2-thinking, were officially discontinued on May 25,… Full reason

    Retired by provider: Moonshot's Kimi API model list states that the kimi-k2 series models, including kimi-k2-thinking, were officially discontinued on May 25, 2026 and are no longer maintained or supported. The weights are still on Hugging Face, but the provider no longer serves the model. Moonshot directs users to kimi-k2.6, kimi-k2.7-code, or its flagship Kimi K3.

  • RemovedDeepSeek-R1
    Retired by provider: DeepSeek-R1 (served as deepseek-reasoner from 20 Jan 2025) is no longer served on DeepSeek's official API: the alias was moved to… Full reason

    Retired by provider: DeepSeek-R1 (served as deepseek-reasoner from 20 Jan 2025) is no longer served on DeepSeek's official API: the alias was moved to V3.1/V3.2 and then V4 thinking mode, and DeepSeek announced on 24 Apr 2026 that the legacy names deepseek-chat/deepseek-reasoner would be discontinued on 24 Jul 2026. The MIT-licensed weights remain downloadable for self-hosting, but the provider no longer serves R1 itself.

  • TierGPT-5.6 TerraA → S
  • TierDeepSeek V4 ProA → S
  • TierGPT-5.4 ProS → A
  • TierGLM-5.1S → A
  • TierGPT-5S → A
  • TierGPT-5.4 miniA → B
  • TierGemini 3 FlashA → B
  • TierMistral LargeA → B
  • TierQwen3.5-397B-A17BA → B
  • TierCommand A+A → B
  • Moved downv0#1 → #4 in Landing Page / Website
  • Moved upLangGraph#5 → #2 in Agents & Automation
  • Moved downKrisp#4 → #11 in Transcription
  • Moved upiZotope RX 12#17 → #5 in Music / Audio
  • Moved downTalkio AI#5 → #13 in Language Learning
  • Moved upMicrosoft Copilot Studio#10 → #5 in Agents & Automation
  • Moved upGPT-5.6 Sol#10 → #5 among all models
  • Moved downZed#5 → #8 in Coding Assistant
  • Moved downGPT-5.4 Pro#6 → #35 among all models
  • Moved downGemini Code Assist#8 → #14 in Coding Assistant
Show 40 more changes

100 smaller rank changes

Catalog refresh: new measurements changed the rank or tier of 18 entries. How scores work

  • RemovedTomeTome Slides discontinued 2025-04-30 (company pivoted to Lightfield CRM).
  • RemovedSupermavenStandalone Supermaven sunset ~2025-11-21 after Cursor (Anysphere) acquisition.
  • TierMiniMax M2.7S → A
  • TierLlama 4S → A
  • TierVrewS → A in Transcription
  • TierScholarcyS → A in Research / Search
  • TierFocal MLA → B in Video
  • TierSlidesPilotA → B in Presentation
  • TierGenra.aiS → B in Video
  • TierAtlas AIS → A in Landing Page / Website
  • TierHoppy CopyS → A in Marketing / Ad Copy
  • TierSpeechyouA → B in Transcription
  • Moved downHoppy Copy#1 → #22 in Marketing / Ad Copy
  • Moved downAtlas AI#1 → #18 in Landing Page / Website
  • Moved downSlidesPilot#3 → #18 in Presentation
  • Moved downHypotenuse AI#3 → #15 in Marketing / Ad Copy
  • Moved downGenra.ai#4 → #20 in Video
  • Moved downPushwoosh ManyMoney AI#4 → #10 in Marketing / Ad Copy
  • Moved downSpeechyou#5 → #26 in Transcription
  • Moved downSpeakly AI#5 → #18 in Language Learning
  • Moved downDeepSite#5 → #17 in Landing Page / Website
  • Moved downScholarcy#5 → #16 in Research / Search
Show 6 more changes
  • TierAnveVoiceS → B in Translation
  • TiercamelAIS → B
  • TierRow64A → B
  • Moved downFocal ML#10 → #19 in Video
  • Moved downLlama 4#13 → #60 among all models
  • Moved downMiniMax M2.7#16 → #48 among all models

1 smaller rank change

Catalog refresh: new measurements changed the rank or tier of 190 entries. How scores work

Show 40 more changes

158 smaller rank changes

Catalog refresh: new measurements changed the rank or tier of 190 entries. How scores work

Show 40 more changes

153 smaller rank changes