Find the right AI for you

Search by name or tell us what you want to do
API servicesChoose how to access modelsAI modelsCapabilities, specifications, and pricingAI toolsFind useful tools by taskMCP directoryConnect tools, data, and workflowsAgent skillsFind reusable skills for AI agentsBenchmarksCompare model performanceOnline toolsConvert accounts, format JSON, estimate costs and test APIsModel timelineExplore model releasesUse casesStart with a practical task
EnterSearch · ↑↓Choose a resultEscClose
Submit中文
MODEL CAPABILITY · PRICE & RELEASES

Capabilities

Artificial Analysis · Compare benchmark results, text costs and release dates separately, with each model and configuration identified.

Data as of2026-10-17 · 273 configurations

Methodology & original report ↗

Capability comparison

One capability index · top configuration per developer · excludes estimates, retired models and future releases.

Chart data · 10 entries
NameCapability index
Claude Opus 5.557.622
Muse Wonder Full55.802
GPT-6 Astra52.674
Gemini 4 Argon52.561
Grok 4.746.447
MiMo-V2.6-Pro46.324
Qwen3.8 Max (0902)45.415
GLM-5.344.777
Step 5 Preview43.734
Kimi K343.594

Capability and text cost

Index ≥30 · cost = (3 × input price + output price) ÷ 4, in USD per million tokens. Showing 60 of 65 configurations, selected by capability.

Chart data · 60 entries
NameCost / capability
Claude Opus 5.5 (Max, Default Fallback)8 / 57.622
Claude Sonnet 5.5 (Max, Default Fallback)4 / 56
Claude Opus 5.5 (Xhigh, Default Fallback)8 / 55.987
Claude Opus 5.5 (High, Default Fallback)8 / 53.583
Claude Fable 5.1 (Max, Default Fallback)20 / 53.355
Claude Fable 5.1 (Xhigh, Default Fallback)20 / 53.203
GPT-6 Astra (Max)20 / 52.674
Gemini 4 Argon (High)4 / 52.561
Muse Wonder (Max)8 / 52.394
GPT-6 Astra (Xhigh)20 / 52.386
Claude Sonnet 5.5 (Xhigh, Default Fallback)4 / 51.896
GPT-6.1 Sol (Max)4 / 51.833
Claude Opus 5.5 (Medium, Default Fallback)8 / 51.243
Claude Fable 5.1 (High, Default Fallback)20 / 51.152
GPT-6.1 Sol (Xhigh)4 / 51.038
GPT-6 Astra (High)20 / 50.919
GPT-6.1 Sol (High)4 / 50.238
GPT-6 Astra (Medium)20 / 49.57
Claude Fable 5.1 (Medium, Default Fallback)20 / 48.921
Muse Spark 1.3 (Max)2 / 48.092
GPT-6.1 Sol (Medium)4 / 47.783
Claude Fable 5.1 (Low, Default Fallback)20 / 46.816
Claude Sonnet 5.5 (High, Default Fallback)4 / 46.753
Grok 4.7 (Xhigh)3 / 46.447
Grok 4.7 (High)3 / 46.332
MiMo-V2.6-Pro0.544 / 46.324
GPT-6 Astra (Low)20 / 45.782
Qwen3.8 Max (0902)3 / 45.415
Muse Spark 1.3 (Xhigh)2 / 45.073
GLM-5.3 (Max)2.15 / 44.777
Step 5 Preview1.425 / 43.734
Kimi K3 (Max)6 / 43.594
Claude Haiku 5.5 (Max)0.2 / 43.395
Claude Opus 5.5 (Low, Default Fallback)8 / 42.308
Grok 4.7 (Low)3 / 42.217
GPT-6.1 Sol (Low)4 / 42.084
GPT-5.6 Terra (Max)4.5 / 42.083
GLM-5.3-Flash0.238 / 41.807
Claude Haiku 5.5 (Xhigh)0.2 / 41.249
Ling 3.1 Flash0.45 / 41.091
Gemini 3.8 Flash (High)1.5 / 40.926
Claude Sonnet 5.5 (Medium, Default Fallback)4 / 40.839
Qwen3.8 2.4T A95B3 / 39.946
Qwen3.8-Flash-Next0.23 / 39.822
Gemini 3.8 Flash (Medium)1.5 / 39.774
DeepSeek V4.1 Flash (Max)0.525 / 39.456
Muse Spark Open (Xhigh)2 / 38.867
Mistral Large 4 Preview2.065 / 38.377
GPT-6 Luna (Max)0.2 / 38.125
GPT-5.6 Terra (Xhigh)4.5 / 37.951
MiMo-V2.6-Flash0.175 / 37.884
Claude Haiku 5.5 (High)0.2 / 37.824
DeepSeek V4 Pro 0813 (Max)1.98 / 35.997
Claude Sonnet 5.5 (Low, Default Fallback)4 / 35.868
DeepSeek V4 Flash Vision (Max)0.66 / 34.839
GPT-6 Luna (Xhigh)0.2 / 34.559
Claude Haiku 5.5 (Medium)0.2 / 34.465
GLM-5.3 (Low)2.15 / 34.299
GPT-5.6 Terra (High)4.5 / 34.238
Qwen3.8 27B (Xhigh)1.125 / 33.676

CapabilitiesDetails

Sorted by capability index, with reasoning configurations kept separate. Estimates and historical records follow rank-eligible entries.

On this page20records
  1. —

    Olmo 3.1 32B Think

    Allen Institute for AI · Estimated score

    Capability index
    7.12
    Input / million tokens
    —
    Output / million tokens
    —
    Value · 3:1 usage mix
    —
  2. —

    Command A

    Cohere · Estimated score

    Capability index
    6.96
    Input / million tokens
    $2.5
    Output / million tokens
    $10
    Value · 3:1 usage mix
    —
  3. —

    Qwen3.5 2B (Reasoning)

    Alibaba · Estimated score

    Capability index
    6.94
    Input / million tokens
    —
    Output / million tokens
    —
    Value · 3:1 usage mix
    —
  4. —

    NVIDIA Nemotron 3 Nano 30B A3B (Non-reasoning)

    NVIDIA · Estimated score

    Capability index
    6.85
    Input / million tokens
    $0.05
    Output / million tokens
    $0.2
    Value · 3:1 usage mix
    —
  5. —

    NVIDIA Nemotron Nano 9B V2 (Non-reasoning)

    NVIDIA · Estimated score

    Capability index
    6.84
    Input / million tokens
    $0.05
    Output / million tokens
    $0.195
    Value · 3:1 usage mix
    —
  6. —

    Hermes 4 - Llama-3.1 70B (Non-reasoning)

    Nous Research · Estimated score

    Capability index
    6.66
    Input / million tokens
    —
    Output / million tokens
    —
    Value · 3:1 usage mix
    —
  7. —

    Sarvam 30B (High)

    Sarvam · Estimated score

    Capability index
    6.56
    Input / million tokens
    —
    Output / million tokens
    —
    Value · 3:1 usage mix
    —
  8. —

    Olmo 3.1 32B Instruct

    Allen Institute for AI · Estimated score

    Capability index
    6.50
    Input / million tokens
    —
    Output / million tokens
    —
    Value · 3:1 usage mix
    —
  9. —

    Gemma 4 E2B (Non-reasoning)

    Google · Estimated score

    Capability index
    6.48
    Input / million tokens
    —
    Output / million tokens
    —
    Value · 3:1 usage mix
    —
  10. —

    R1 1776

    Perplexity · Estimated score

    Capability index
    6.44
    Input / million tokens
    —
    Output / million tokens
    —
    Value · 3:1 usage mix
    —
  11. —

    Llama 3.2 Instruct 90B (Vision)

    Meta · Estimated score

    Capability index
    6.41
    Input / million tokens
    —
    Output / million tokens
    —
    Value · 3:1 usage mix
    —
  12. —

    Phi-4 Mini Instruct

    Microsoft · Estimated score

    Capability index
    6.33
    Input / million tokens
    $0.075
    Output / million tokens
    $0.3
    Value · 3:1 usage mix
    —
  13. —

    EXAONE 4.0 32B (Non-reasoning)

    LG AI Research · Estimated score

    Capability index
    6.33
    Input / million tokens
    —
    Output / million tokens
    —
    Value · 3:1 usage mix
    —
  14. —

    Qwen3.5 2B (Non-reasoning)

    Alibaba · Estimated score

    Capability index
    6.19
    Input / million tokens
    —
    Output / million tokens
    —
    Value · 3:1 usage mix
    —
  15. —

    Qwen3.5 0.8B (Reasoning)

    Alibaba · Estimated score

    Capability index
    6.13
    Input / million tokens
    —
    Output / million tokens
    —
    Value · 3:1 usage mix
    —
  16. —

    DeepHermes 3 - Mistral 24B Preview (Non-reasoning)

    Nous Research · Estimated score

    Capability index
    6.07
    Input / million tokens
    —
    Output / million tokens
    —
    Value · 3:1 usage mix
    —
  17. —

    Jamba 1.7 Large

    AI21 Labs · Estimated score

    Capability index
    6.07
    Input / million tokens
    —
    Output / million tokens
    —
    Value · 3:1 usage mix
    —
  18. —

    Granite 4.0 H Small

    IBM · Estimated score

    Capability index
    6.05
    Input / million tokens
    —
    Output / million tokens
    —
    Value · 3:1 usage mix
    —
  19. —

    Qwen3 Omni 30B A3B Instruct

    Alibaba · Estimated score

    Capability index
    6.01
    Input / million tokens
    $0.25
    Output / million tokens
    $0.97
    Value · 3:1 usage mix
    —
  20. —

    LFM2 24B A2B

    Liquid AI · Estimated score

    Capability index
    5.95
    Input / million tokens
    —
    Output / million tokens
    —
    Value · 3:1 usage mix
    —
Ranking criteria & test conditions
  • Intelligence scores come from the same leaderboard snapshot. Reasoning configurations remain separate; estimated scores and retired models are excluded from the highlights.
  • Release dates follow the report and are separate from our update dates. Undated models are excluded from recent releases. A report’s date is not treated as independent proof of an official model release.
  • Text prices use USD per million tokens, with input and output stored separately. Subscription, cache and media charges are not mixed into this comparison.
  • Value picks require an intelligence score of at least 30 and rank by score ÷ (3 × input price + output price) × 4, selecting the highest value for each developer. This assumes a 3:1 text input/output ratio; actual task costs vary.
  • Developer highlights show one representative per developer. Model aliases and reasoning configurations are not automatically mapped to a specific catalog version. Full results retain the original configurations and missing values.

Charts are calculated from the full report, independently of the detail page. Release activity uses release dates, not changes in model capability over time. Missing prices are not zero costs.