Skip to contentT/TTokens or Towers

Compare

LeaderboardEvery tracked model, ranked and priced side by sideBenchmarksWhat each test measures and who leads itPricesAPI, media, and hardware rates with a cost calculator

Plan

Stack plannerSix questions to a priced local and hosted stackShortlistOur picks for subscriptions, APIs, and hardwareEnterpriseSeat plans against self-hosting for a whole team

Build

HardwareGPUs and workstations, and which models they fitHostsWhere to rent open-weight inference, and for how muchHarnessesCoding agents and the models they drive

Lab

FindingsResults from our own evaluationsScanAudit a repo’s AI usage in the browserMethodWhere the numbers come from and how we check them
Data checked September 25, 2026
Data checked September 25, 2026
Plan a stack
T/TTokens or Towers

An independent guide to AI models, what they cost, and the hardware to run them. No account, no API key, no sponsors.

Compare

  • Leaderboard
  • Benchmarks
  • Prices

Plan

  • Stack planner
  • Shortlist
  • Enterprise

Build

  • Hardware
  • Hosts
  • Harnesses

Lab

  • Findings
  • Scan
  • Method
Data checked September 25, 2026 · Every number links to its sourceRSS · Source on GitHub
Leaderboard/Meta
Meta

Meta · Llama 4

Llama 4 Maverick

Open weightsSupersededLlama 4 Community License

Meta's 2025 open-weights MoE model (Llama 4 Community License); superseded by the Muse line and no longer served on a Meta API.

AA Intelligence
10#99 of 142
Price
Self-host
Output speed
115tok/s
Context window
1M tokens

Composite indexes

  • Artificial Analysis Intelligence Index#99 of 142
    10independentAA Intelligence Index v4.3.2, non-reasoningArtificial Analysis: Llama 4 Maverick ↗

    Leader: Claude Opus 5.5 at 58

Coding

  • Terminal-Bench 4.0#63 of 88
    0%independentTerminal-Bench 4.0 (AA harness), non-reasoningArtificial Analysis: Llama 4 Maverick ↗

    Leader: Claude Mythos 5.1 at 60.9%

  • Terminal-Bench 2.x#93 of 103
    7.9%independentTerminal-Bench 2.1 (AA harness), non-reasoningArtificial Analysis: Llama 4 Maverick ↗

    Leader: Claude Fable 5.1 at 91.4%

  • SciCode#86 of 94
    31.7%independentnon-reasoningArtificial Analysis: Llama 4 Maverick ↗

    Leader: Claude Opus 5.5 at 66.9%

  • LiveCodeBench#39 of 56
    39.7%independentAA, non-reasoningArtificial Analysis: Llama 4 Maverick ↗

    Leader: Fugu at 92.9%

Agents and tool use

  • GDPval-AA#102 of 103
    -291independentGDPval-AA, non-reasoningArtificial Analysis: Llama 4 Maverick ↗

    Leader: Claude Opus 5.5 at 1846

  • τ²-Bench#68 of 88
    17.8%independenttelecom (AA), non-reasoningArtificial Analysis: Llama 4 Maverick ↗

    Leader: GLM-5.2 at 99.1%

Reasoning and knowledge

  • Humanity’s Last Exam#123 of 147
    4.9%independentno tools (AA), non-reasoningArtificial Analysis: Llama 4 Maverick ↗

    Leader: Claude Opus 5.5 at 61.4%

  • GPQA Diamond#104 of 142
    67.1%independentAA, non-reasoningArtificial Analysis: Llama 4 Maverick ↗

    Leader: GPT-6 Astra at 96.1%

  • CritPt#91 of 139
    0%independentnon-reasoningArtificial Analysis: Llama 4 Maverick ↗

    Leader: GPT-5.6 Sol at 32.3%

  • ARC-AGI-2#24 of 25
    0%independentsemi-private,ARC Prize leaderboard ↗

    Leader: GPT-6 Astra at 95%

Vision and long context

  • AA Long Context Reasoning#91 of 139
    50%independentnon-reasoningArtificial Analysis: Llama 4 Maverick ↗

    Leader: Kimi K3 at 88.7%

  • MMMU-Pro#56 of 72
    62.1%independentAA, non-reasoningArtificial Analysis: Llama 4 Maverick ↗

    Leader: Claude Opus 5.5 at 87.7%

Human preference

  • LMArena Text#52 of 54
    1327independentllama-4-maverick, style control default, vote cutoff 2026-09-13LMArena Text leaderboard ↗

    Leader: Claude Fable 5 at 1506

Other published results

ResultScoreReported bySource
aime-2025 · AA, non-reasoning19.3%independentArtificial Analysis: Llama 4 Maverick ↗
AA-Omniscience Index · non-reasoning-41.9 index (-100 to 100)independentArtificial Analysis: Llama 4 Maverick ↗
AutomationBench-AA (partial score) · non-reasoning0.2%independentArtificial Analysis: Llama 4 Maverick ↗

Specs

Released
April 5, 2025
Parameters
402B total, 17B active
Size class
100 to 500B
Input
text, image
Output
text
Context
1M tokens
Max output
Not published
First token
0.90s
  • Speed: Artificial Analysis: Llama 4 Maverick (median output speed; TTFT is time to first chunk incl. reasoning) ↗

Will it fit your budget?

The planner prices this model against local hardware and other APIs for your workload.

Plan a stack →

Compare with

  • CohereNorth Mini Code 1.09.9
  • NVIDIANemotron 3 Nano Omni 30B A3B Reasoning10.3
  • GoogleDiffusionGemma 26B A4B9.5
  • Prime IntellectINTELLECT-310.6