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
1367 models808 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

math benchmark

SciCode Leaderboard

SciCode is a research coding benchmark with 338 subproblems from 80 scientific problems across 16 natural science sub-fields.

Updated Oct 23, 2026

Models31
Model coverage31
MetricScore
EvidenceB

On this page

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

SciCode Ranking

Higher score ranks better on this benchmark.

30 of 31 rows
Columns

Show columns

Sort by
Rank
Model
Score
Percentile
Participants
Evidence
Evaluated
Rank01ModelMAMistral Large 4Mistral AIScore91.80%Percentile100.00%Participants31EvidenceCEvaluatedOct 22, 2026
Rank02ModelBYSeed 2.1 ProByteDanceScore59.80%Percentile96.67%Participants31EvidenceCEvaluatedOct 22, 2026
Rank03ModelGOGemini 3.1 ProGoogleScore59.00%Percentile93.33%Participants31EvidenceCEvaluatedOct 22, 2026
Rank04ModelMEMuse Spark OpenMetaScore58.50%Percentile90.00%Participants31EvidenceCEvaluatedOct 22, 2026
Rank05ModelBYSeed 2.1 TurboByteDanceScore57.80%Percentile86.67%Participants31EvidenceCEvaluatedOct 22, 2026
Rank06ModelACQwen3.7 MaxAlibaba Cloud / Qwen TeamScore53.50%Percentile83.33%Participants31EvidenceCEvaluatedOct 22, 2026
Rank07ModelMAKimi K2.6Moonshot AIScore52.20%Percentile80.00%Participants31EvidenceCEvaluatedOct 22, 2026
Rank08ModelACQwen3.7-PlusAlibaba Cloud / Qwen TeamScore51.30%Percentile76.67%Participants31EvidenceCEvaluatedOct 22, 2026
Rank09ModelRABeam 1.0Reflection AIScore49.20%Percentile73.33%Participants31EvidenceCEvaluatedOct 22, 2026
Rank10ModelTMInkling-SmallThinking Machines LabScore48.70%Percentile70.00%Participants31EvidenceCEvaluatedOct 22, 2026
Rank11ModelMAKimi K2.5Moonshot AIScore48.70%Percentile66.67%Participants31EvidenceCEvaluatedOct 22, 2026
Rank12ModelUPSolar Mini 4UpstageScore47.60%Percentile63.33%Participants31EvidenceBEvaluatedOct 22, 2026
Rank13ModelNVNemotron 3.5 Super VL (120B A12B)NVIDIAScore47.13%Percentile60.00%Participants31EvidenceCEvaluatedOct 22, 2026
Rank14ModelMAKimi K2-Thinking-0905Moonshot AIScore44.80%Percentile56.67%Participants31EvidenceCEvaluatedOct 22, 2026
Rank15ModelNVNemotron 3 Ultra (550B A55B)NVIDIAScore44.60%Percentile53.33%Participants31EvidenceCEvaluatedOct 22, 2026
Rank16ModelMEMuse Glimmer-30BMetaScore43.60%Percentile50.00%Participants31EvidenceCEvaluatedOct 22, 2026
Rank17ModelNVNemotron 3 Super (120B A12B)NVIDIAScore42.05%Percentile46.67%Participants31EvidenceCEvaluatedOct 22, 2026
Rank18ModelZAGLM-4.5Zhipu AIScore41.70%Percentile43.33%Participants31EvidenceCEvaluatedOct 22, 2026
Rank19ModelNVNemotron 3.5 Bolt (74B A9B 24GB)NVIDIAScore41.02%Percentile40.00%Participants31EvidenceCEvaluatedOct 22, 2026
Rank20ModelMIMiniMax M2.1MiniMaxScore39.00%Percentile36.67%Participants31EvidenceCEvaluatedOct 22, 2026
Rank21ModelIBIBM Granite 4.2 30BIBMScore38.76%Percentile33.33%Participants31EvidenceCEvaluatedOct 22, 2026
Rank22ModelCONorth Mini Code 1.0CohereScore38.20%Percentile30.00%Participants31EvidenceCEvaluatedOct 22, 2026
Rank23ModelCOCommand A+CohereScore38.00%Percentile26.67%Participants31EvidenceCEvaluatedOct 22, 2026
Rank24ModelINMercury 2InceptionScore38.00%Percentile23.33%Participants31EvidenceCEvaluatedOct 22, 2026
Rank25ModelZAGLM-4.5-AirZhipu AIScore37.30%Percentile20.00%Participants31EvidenceCEvaluatedOct 22, 2026
Rank26ModelIBIBM Granite 4.2 8BIBMScore36.09%Percentile16.67%Participants31EvidenceCEvaluatedOct 22, 2026
Rank27ModelMIMiniMax M2MiniMaxScore36.00%Percentile13.33%Participants31EvidenceCEvaluatedOct 22, 2026
Rank28ModelNVNemotron Labs Dory 1.0 (4B)NVIDIAScore34.94%Percentile10.00%Participants31EvidenceCEvaluatedOct 22, 2026
Rank29ModelNVNemotron 3 Nano (30B A3B)NVIDIAScore33.30%Percentile6.67%Participants31EvidenceCEvaluatedOct 22, 2026
Rank30ModelNVNemotron 3.5 Lightning (30B A3B)NVIDIAScore32.60%Percentile3.33%Participants31EvidenceCEvaluatedOct 22, 2026

SciCode Highlights

The leading models and scores on this benchmark.

SciCode Score Distribution

A closer view of the leading scores on this benchmark.

SciCode

The Top AI Models for SciCode

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

What is SciCode?

What SciCode measures and how its scores work.

SciCode challenges language models to code solutions for scientific problems, including mathematics, physics, chemistry, biology, and materials science.

It measures performance involving knowledge recall, reasoning, and code synthesis, reported as Score in ratio units.

Scores are shown in ratio. This benchmark is not independently verified and has an evidence level of B.

Family
SciCode
Modality
text
Primary category
math
Score direction
higher
LLMBoard eligible
No
Evaluation key
overall

LLM Stats. Benchmark scores retain their original unit. Overall score eligibility is shown separately.

FAQ

Common questions about SciCode.

Which model scores highest on SciCode?

Mistral Large 4 is currently ranked first with 91.80%.

What are the top three models on SciCode?

The current leaders are Mistral Large 4 (91.80%), Seed 2.1 Pro (59.80%), and Gemini 3.1 Pro (59.00%).

Which SciCode model has the lowest official input price?

Solar Mini 4 has the lowest matched official input price at $0.10 input / $0.40 output per 1M tokens.

Which models are fastest among SciCode results?

The fastest matched records are Mercury 2 (1,009.00 tok/s via Inception), MiniMax M2.1 (100.00 tok/s via MiniMax), and Gemini 3.1 Pro (90.00 tok/s via Google).

Does the highest score 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 SciCode measure?

It measures performance involving knowledge recall, reasoning, and code synthesis, reported as Score in ratio units.

Is a higher score better?

Yes. Higher values rank better for this benchmark.

How many models are compared?

31 model results are currently shown.

Does this benchmark affect the overall score?

No. This benchmark is shown for reference but does not contribute to the overall score.

Rank #1Mistral Large 491.80%
Rank #2Seed 2.1 Pro59.80%
Rank #3Gemini 3.1 Pro59.00%
Rank #4Muse Spark Open58.50%

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

  1. 01
    MA
    Mistral Large 4Mistral AI
    Score
    91.80%
    Price
    $1.4 input / $4.2 output per 1M tokens

    Strengths

    • Ranks #1 of 31 compared models
    • 100th percentile on this benchmark
    • C evidence result

    Considerations

    • This result measures SciCode, not total model capability
  2. 02
    BY
    Seed 2.1 ProByteDance
    Score
    59.80%

    Strengths

    • Ranks #2 of 31 compared models
    • 97th percentile on this benchmark
    • C evidence result

    Considerations

    • This result measures SciCode, not total model capability
  3. 03
    GO
    Google
    Score
    59.00%
    Price
    $2.0 input / $12 output per 1M tokens
    Speed
    Up to 90.00 tok/s via Google

    Strengths

    • Ranks #3 of 31 compared models
    • 93th percentile on this benchmark
    • C evidence result

    Considerations

    • This result measures SciCode, not total model capability
  4. 04
    ME
    Meta
    Score
    58.50%

    Strengths

    • Ranks #4 of 31 compared models
    • 90th percentile on this benchmark
    • C evidence result

    Considerations

    • This result measures SciCode, not total model capability
  5. 05
    BY
    ByteDance
    Score
    57.80%

    Strengths

    • Ranks #5 of 31 compared models
    • 87th percentile on this benchmark
    • C evidence result

    Considerations

    • This result measures SciCode, not total model capability

Selection summary

Best AI Models for SciCode

Mistral Large 4 currently leads SciCode with 91.80%. 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 3.1 Pro
Muse Spark Open
Seed 2.1 Turbo
Benchmark rank #1Mistral Large 491.80% · $1.4 input / $4.2 output per 1M tokens
Benchmark rank #2Seed 2.1 Pro59.80%
Benchmark rank #3Gemini 3.1 Pro59.00% · $2.0 input / $12 output per 1M tokens