The AI Times
EN
Volume I · Daily editionDaily roundup of AI newsMCPMD
EN
THE AI TIMES

Models / benchmarks

Benchmarks

Intelligence IndexHigher is better

AnthropicOpenAIGoogleOthers
  1. 65
    Claude Fable 5.2 (Max, Default Fallback)
  2. 62
    Claude Fable 5.2 (Xhigh, Default Fallback)
  3. 62
    Claude Fable 5.2 (High, Default Fallback)
  4. 60
    Gemini 4 Carbon (High)
  5. 59
    Claude Opus 5.5 (Max, Default Fallback)
  6. 58
    Gemini 4 Carbon (Medium)
  7. 57
    Claude Opus 5.5 (Xhigh, Default Fallback)
  8. 57
    GPT-6 Astra (Max)
  9. 56
    Claude Fable 5.2 (Medium, Default Fallback)
  10. 55
    GPT-6 Astra (Xhigh)
  11. 55
    GPT-6 Astra (High)
  12. 55
    GPT-6.1 Sol (Max)
  13. 55
    Claude Sonnet 5.5 (Max, Default Fallback)
  14. 54
    Gemini 4 Carbon (Low)
  15. 54
    Claude Fable 5.2 (Low, Default Fallback)
  16. 53
    GPT-6.1 Sol (Xhigh)
  17. 52
    Claude Opus 5.5 (High, Default Fallback)
  18. 52
    GPT-6 Astra (Medium)
  19. 51
    GPT-6.1 Terra (Max)
  20. 50
    Gemini 4 Argon (High)
  21. 50
    GPT-6.1 Sol (High)
  22. 50
    Muse Wonder Full (Max)
  23. 50
    Claude Fable 5.1 (Xhigh, Default Fallback)
  24. 50
    Claude Sonnet 5.5 (Xhigh, Default Fallback)
  25. 49
    Claude Fable 5.1 (Max, Default Fallback)
  26. 48
    Claude Opus 5.5 (Medium, Default Fallback)
  27. 47
    Claude Fable 5.1 (High, Default Fallback)
  28. 47
    Kimi K3.1 Pro
  29. 46
    GPT-6.1 Sol (Medium)
  30. 45
    GPT-6.1 Terra (Xhigh)

Full comparison at Artificial Analysis →
Updated: 12 November 2026

Image generation

Text to Image Arena · Elo

  1. GPT Image 2.6 Sunburst (max)1,262
  2. GPT Image 2.5 Sunburst (max)1,230
  3. GPT Image 2.6 Flare (max)1,227
  4. GPT Image 2.5 Flare (max)1,211
  5. GPT Image 2 (high)1,189
  6. Nano Banana 2.11,177
  7. MAI-Image-2.61,170
  8. Grok Imagine Image 2.01,167
  9. Nano Banana 2 (Gemini 3.1 Flash Image)1,144
  10. Seedream 5.51,135

The axis starts at 1,100 Elo.

Full comparison at Artificial Analysis →
Updated: 12 November 2026

FrontierCode 2.0

Measures functional correctness, interpretation of user intent, and the design and integration of changes into existing code; the score is a weighted summary of the criteria.

  1. Claude Opus 5.545.5 %
  2. GPT-6 Astra42.8 %
  3. Grok 4.728.6 %
  4. Gemini 4 Argon26.3 %
  5. Muse Wonder 1.024.3 %
  6. SWE-223.9 %
  7. DeepSeek V4.1 Flash19.3 %
  8. Kimi K317.7 %

Full comparison at Cognition →
Updated: 12 November 2026

Model speed

Output tokens per second · higher is faster. Time to first answer token is shown in seconds directly from the source.

  1. Gemini 4 Carbon (High)99.5 tokens/s

    First answer token: 125.2 s

  2. Claude Opus 5.5 (Max, Default Fallback)97.3 tokens/s

    First answer token: 770.1 s

  3. Gemini 4 Carbon (Medium)93.6 tokens/s

    First answer token: 27.5 s

  4. Claude Fable 5.2 (Max, Default Fallback)92.5 tokens/s

    First answer token: 242.0 s

  5. Claude Fable 5.2 (Xhigh, Default Fallback)86.0 tokens/s

    First answer token: 97.3 s

  6. Claude Opus 5.5 (Xhigh, Default Fallback)81.4 tokens/s

    First answer token: 171.1 s

  7. Claude Fable 5.2 (Medium, Default Fallback)81.0 tokens/s

    First answer token: 18.3 s

  8. Claude Fable 5.2 (High, Default Fallback)80.4 tokens/s

    First answer token: 30.2 s

  9. GPT-6 Astra (Xhigh)50.0 tokens/s

    First answer token: 325.2 s

  10. GPT-6 Astra (Max)46.4 tokens/s

    First answer token: 311.9 s

Full comparison at Artificial Analysis →
Updated: 12 November 2026

Price per 1 million tokens

USD · blended price from the source. A shorter bar means a lower price. Model settings are quoted verbatim.

  1. GPT-6 Astra (Max)20 $

    Input 10 $ · output 50 $

  2. GPT-6 Astra (Xhigh)20 $

    Input 10 $ · output 50 $

  3. Claude Fable 5.2 (Max, Default Fallback)16 $

    Input 8 $ · output 40 $

  4. Claude Fable 5.2 (Xhigh, Default Fallback)16 $

    Input 8 $ · output 40 $

  5. Claude Fable 5.2 (High, Default Fallback)16 $

    Input 8 $ · output 40 $

  6. Claude Fable 5.2 (Medium, Default Fallback)16 $

    Input 8 $ · output 40 $

  7. Gemini 4 Carbon (High)8 $

    Input 4 $ · output 20 $

  8. Claude Opus 5.5 (Max, Default Fallback)8 $

    Input 4 $ · output 20 $

  9. Gemini 4 Carbon (Medium)8 $

    Input 4 $ · output 20 $

  10. Claude Opus 5.5 (Xhigh, Default Fallback)8 $

    Input 4 $ · output 20 $

Full comparison at Artificial Analysis →
Updated: 12 November 2026

Subscription value

Monthly value at API prices · a shared dollar scale for all plans.

20 $ / month

  1. GPT-6.1 Sol211 $ / month

    $20 · 1.0 billion tokens / month

  2. Opus 5.51,178 $ / month

    $20 · 2.9 billion tokens / month

100 $ / month

  1. GPT-6.1 Sol1,055 $ / month

    $100 · 5.1 billion tokens / month

  2. Opus 5.55,725 $ / month

    $100 · 14.0 billion tokens / month

200 $ / month

  1. GPT-6.1 Sol2,084 $ / month

    $200 · 10.2 billion tokens / month · 10.4× monthly fee

  2. Opus 5.511,726 $ / month

    $200 · 28.6 billion tokens / month · 58.6× monthly fee

SemiAnalysis describes experiments that isolate token types, observe complete usage-meter steps, estimate rates for each usage window, and blend them into workload scenarios before applying API list prices. Source methodology [1].

Full comparison at tokenusage.dev →
Updated: 12 November 2026