llmboard.aiAI model intelligence
Home

Model Rankings

OverallOpen ModelsAgentCodingReasoningMathKnowledgeInstruction FollowingTextVision
Image GenerationImage Editing
Video GenerationImage to VideoVideo Editing
Text to SpeechSpeech to Text
Embeddings

Efficiency

Chat Token PricingImage PricingVideo PricingAudio Pricing
Chat Speed & LatencyProvider Reliability

Benchmarks

GPQAMMLU-ProAIME 2025SWE-Bench VerifiedMMLUHumanity's Last ExamLiveCodeBenchMATHHumanEvalMMMU-Pro
All Benchmarks

Tools

Model Directory

Scoring & Data

Scoring & Data
1281 models767 benchmarks

Leaderboard Center

Overall RankingCodingCore BenchmarksPrice & ValueRuntime Performance

Modalities

All ModelsImage GenerationImage EditingVideo GenerationImage-to-VideoVideo EditingText-to-SpeechSpeech-to-TextEmbeddings

Data & Methods

Scoring MethodAll Benchmarks
llmboard.aiCopyright 2026 llmboard.ai

coding tools benchmark

AA SciCode Subtasks Leaderboard

288 subtasks with background; v1.0.1; 3 runs; pass@1; no model tools; isolated execution; 300 seconds

Updated Sep 22, 2026

Models89
Model coverage89
Metricscicode
EvidenceB

On this page

  • Ranking
  • Highlights
  • Distribution
  • Top models
  • About
  • FAQ

AA SciCode Subtasks Ranking

Higher scicode ranks better on this benchmark.

30 of 89 rows
Columns

Show columns

Sort by
Rank
Model
scicode
Percentile
Participants
Evidence
Evaluated
Rank01ModelANClaude Fable 5.1Anthropicscicode63.08%Percentile100.00%Participants89EvidenceBEvaluatedN/A
Rank02ModelANClaude Fable 5Anthropicscicode61.00%Percentile98.86%Participants89EvidenceBEvaluatedN/A
Rank03ModelGOGemini 3.7 FlashGooglescicode59.84%Percentile97.73%Participants89EvidenceBEvaluatedN/A
Rank04ModelMEMuse Spark 1.3Metascicode59.72%Percentile96.59%Participants89EvidenceBEvaluatedN/A
Rank05ModelMAKimi K3Moonshot AIscicode59.49%Percentile95.45%Participants89EvidenceBEvaluatedN/A
Rank06ModelZAGLM-5.3Zhipu AIscicode59.03%Percentile94.32%Participants89EvidenceBEvaluatedN/A
Rank07ModelMEMuse Spark 1.1Metascicode58.80%Percentile93.18%Participants89EvidenceBEvaluatedN/A
Rank08ModelGOGemini 3.1 ProGooglescicode58.68%Percentile92.05%Participants89EvidenceBEvaluatedN/A
Rank09ModelOPGPT-5.6 SolOpenAIscicode57.75%Percentile90.91%Participants89EvidenceBEvaluatedN/A
Rank10ModelMEMuse Spark 1.2Metascicode57.41%Percentile89.77%Participants89EvidenceBEvaluatedN/A
Rank11ModelGOGemini 3.8 FlashGooglescicode56.60%Percentile88.64%Participants89EvidenceBEvaluatedN/A
Rank12ModelXAGrok 4.6xAIscicode56.48%Percentile87.50%Participants89EvidenceBEvaluatedN/A
Rank13ModelOPGPT-6 AstraOpenAIscicode56.48%Percentile86.36%Participants89EvidenceBEvaluatedN/A
Rank14ModelANClaude Opus 5Anthropicscicode56.37%Percentile85.23%Participants89EvidenceBEvaluatedN/A
Rank15ModelOPGPT-5.5OpenAIscicode56.13%Percentile84.09%Participants89EvidenceBEvaluatedN/A
Rank16ModelOPGPT-5.6 TerraOpenAIscicode54.98%Percentile82.95%Participants89EvidenceBEvaluatedN/A
Rank17ModelXAGrok 4.5xAIscicode54.98%Percentile81.82%Participants89EvidenceBEvaluatedN/A
Rank18ModelANClaude Opus 4.8Anthropicscicode54.40%Percentile80.68%Participants89EvidenceBEvaluatedN/A
Rank19ModelANClaude Sonnet 5Anthropicscicode54.28%Percentile79.55%Participants89EvidenceBEvaluatedN/A
Rank20ModelGOGemini 3.5 FlashGooglescicode53.94%Percentile78.41%Participants89EvidenceBEvaluatedN/A
Rank21ModelOPGPT-5.6 LunaOpenAIscicode53.59%Percentile77.27%Participants89EvidenceBEvaluatedN/A
Rank22ModelGOGemini 3.6 FlashGooglescicode53.36%Percentile76.14%Participants89EvidenceBEvaluatedN/A
Rank23ModelACQwen3.8 MaxAlibaba Cloud / Qwen Teamscicode53.24%Percentile75.00%Participants89EvidenceBEvaluatedN/A
Rank24ModelOPGPT-5.4 miniOpenAIscicode52.08%Percentile73.86%Participants89EvidenceBEvaluatedN/A
Rank25ModelDEDeepSeek-V4.1-FlashDeepSeekscicode51.85%Percentile72.73%Participants89EvidenceBEvaluatedN/A
Rank26ModelZAGLM-5.3-FlashZhipu AIscicode51.62%Percentile71.59%Participants89EvidenceBEvaluatedN/A
Rank27ModelMAKimi K2.6Moonshot AIscicode51.50%Percentile70.45%Participants89EvidenceBEvaluatedN/A
Rank28ModelZAGLM-5.2Zhipu AIscicode51.16%Percentile69.32%Participants89EvidenceBEvaluatedN/A
Rank29ModelDEDeepSeek-V4-Pro-0813DeepSeekscicode51.04%Percentile68.18%Participants89EvidenceBEvaluatedN/A
Rank30ModelXIMiMo-V2.5-ProXiaomiscicode50.58%Percentile67.05%Participants89EvidenceBEvaluatedN/A

AA SciCode Subtasks Highlights

The leading models and scores on this benchmark.

AA SciCode Subtasks Score Distribution

A closer view of the leading scores on this benchmark.

AA SciCode Subtasks

The Top AI Models for AA SciCode Subtasks

The first five results on this benchmark, with official price and output speed added where the model identity can be matched.

What is AA SciCode Subtasks?

What AA SciCode Subtasks measures and how its scores work.

288 subtasks with background; v1.0.1; 3 runs; pass@1; no model tools; isolated execution; 300 seconds

Scores are shown in ratio. This benchmark is verified and has an evidence level of B.

Family
scicode
Modality
text
Primary category
coding tools
Score direction
higher
LLMBoard eligible
Yes
Evaluation key
overall

Artificial Analysis. Benchmark scores retain their original unit. Overall score eligibility is shown separately.

FAQ

Common questions about AA SciCode Subtasks.

Which model scores highest on AA SciCode Subtasks?

Claude Fable 5.1 is currently ranked first with 63.08%.

What are the top three models on AA SciCode Subtasks?

The current leaders are Claude Fable 5.1 (63.08%), Claude Fable 5 (61.00%), and Gemini 3.7 Flash (59.84%).

Which AA SciCode Subtasks model has the lowest official input price?

MiniStral 3 (14B Instruct 2512) has the lowest matched official input price at $0.10 input / $0.10 output per 1M tokens.

Which models are fastest among AA SciCode Subtasks results?

The fastest matched records are Mercury 2 (1,009.00 tok/s via Inception), GPT OSS 20B (1,000.00 tok/s via Groq), and Llama 4 Scout (776.10 tok/s via Groq).

Does the highest scicode result prove overall model quality?

No. This benchmark measures one defined capability or task. The overall LLMBoard score uses a separate aggregation across eligible benchmark evidence.

What does AA SciCode Subtasks measure?

288 subtasks with background; v1.0.1; 3 runs; pass@1; no model tools; isolated execution; 300 seconds

Is a higher scicode better?

Yes. Higher values rank better for this benchmark.

How many models are compared?

89 model results are currently shown.

Does this benchmark affect the overall score?

Yes. This benchmark can contribute to the current LLMBoard capability score.

Rank #1Claude Fable 5.163.08%
Rank #2Claude Fable 561.00%
Rank #3Gemini 3.7 Flash59.84%
Rank #4Muse Spark 1.359.72%

Ranking basisThis aa scicode subtasks AI model leaderboard uses descending scicode in the benchmark original unit. The leaderboard ranking keeps matched price and speed data separate from benchmark evidence.

  1. 01
    AN
    Claude Fable 5.1Anthropic
    scicode
    63.08%
    Price
    $10 input / $50 output per 1M tokens
    Speed
    Up to 3.94 tok/s via Anthropic

    Strengths

    • Ranks #1 of 89 compared models
    • 100th percentile on this benchmark
    • B evidence result

    Considerations

    • This result measures AA SciCode Subtasks, not total model capability
  2. 02
    AN
    Claude Fable 5Anthropic
    scicode
    61.00%
    Price
    $10 input / $50 output per 1M tokens

    Strengths

    • Ranks #2 of 89 compared models
    • 99th percentile on this benchmark
    • B evidence result

    Considerations

    • This result measures AA SciCode Subtasks, not total model capability
  3. 03
    GO
    Google
    scicode
    59.84%
    Price
    $0.75 input / $3.8 output per 1M tokens
    Speed
    Up to 42.46 tok/s via Google

    Strengths

    • Ranks #3 of 89 compared models
    • 98th percentile on this benchmark
    • B evidence result

    Considerations

    • This result measures AA SciCode Subtasks, not total model capability
  4. 04
    ME
    Meta
    scicode
    59.72%
    Price
    $1.3 input / $4.3 output per 1M tokens
    Speed
    Up to 4.64 tok/s via Meta Model API

    Strengths

    • Ranks #4 of 89 compared models
    • 97th percentile on this benchmark
    • B evidence result

    Considerations

    • This result measures AA SciCode Subtasks, not total model capability
  5. 05
    MA
    Moonshot AI
    scicode
    59.49%
    Price
    $3.0 input / $15 output per 1M tokens
    Speed
    Up to 4.03 tok/s via Moonshot AI

    Strengths

    • Ranks #5 of 89 compared models
    • 95th percentile on this benchmark
    • B evidence result

    Considerations

    • This result measures AA SciCode Subtasks, not total model capability

Selection summary

Best AI Models for AA SciCode Subtasks

Claude Fable 5.1 currently leads AA SciCode Subtasks with 63.08%. It is the top model on this specific benchmark, while the best LLM for the broader task should also be checked against other benchmarks, price, and runtime.

Use this leaderboard with the supporting benchmark results and coverage details above. A leaderboard position summarizes the selected ranking signal; it does not replace workload-specific testing.

Gemini 3.7 Flash
Muse Spark 1.3
Kimi K3
Benchmark rank #1Claude Fable 5.163.08% · $10 input / $50 output per 1M tokens
Benchmark rank #2Claude Fable 561.00% · $10 input / $50 output per 1M tokens
Benchmark rank #3Gemini 3.7 Flash59.84% · $0.75 input / $3.8 output per 1M tokens