CodingActive
Trust Score:
93

SciCode

Research-level scientific coding benchmark: models write code to solve subproblems drawn from real scientific workflows across natural-science domains.

Launched: 2024-07-18Refresh: static
Status Assessment (active):

Actively evaluated by Artificial Analysis on current frontier models; no saturation evidence recorded.

Performance Timeline

Longitudinal progression of model scores against human baselines.

Performance & Historical Trajectory

Empirical score progression across model release dates and evaluation rounds.

Independent Vendor-Reported
ModelScoreDateSource TypeProvenance
A.X-K241%2026-11-15independentSource ↗
Llama 4 Scout21.3%2026-11-15independentSource ↗
Qwen3.7 Plus46.1%2026-11-15independentSource ↗
Ling 3.1 Flash54.1%2026-11-15independentSource ↗
NVIDIA Nemotron 3 Nano 4B16.1%2026-11-15independentSource ↗
K2 Horizon 0.9B6.9%2026-11-15independentSource ↗
Qwen3.8-Flash-Next50.6%2026-11-15independentSource ↗
Ministral 3 8B20.7%2026-11-15independentSource ↗
Llama 4 Maverick31.7%2026-11-15independentSource ↗
Qwen3.8 2.4T A95B54.1%2026-11-15independentSource ↗
Kimi K2.7 Code47.8%2026-11-15independentSource ↗
DeepSeek V4-Pro51%2026-11-15independentSource ↗
Granite 4.2 8B31.5%2026-11-15independentSource ↗
Celeris-121.6%2026-11-15independentSource ↗
Ling 3.0 Tiny24.2%2026-11-15independentSource ↗
Solar Open2 250B48%2026-11-15independentSource ↗
Nemotron 3.5 Bolt 74B-A9B39.7%2026-11-15independentSource ↗
DeepSeek V4 Flash Vision49.7%2026-11-15independentSource ↗
Mercury 2.538.5%2026-11-15independentSource ↗
Gemini 3.8 Flash56.6%2026-11-15independentSource ↗
Inkling Small49.7%2026-11-15independentSource ↗
Ministral 3 14B23.8%2026-11-15independentSource ↗
Gemma 4 26B A4B40%2026-11-15independentSource ↗
Gemma 4 31B45.5%2026-11-15independentSource ↗
Qwen3 Coder Next36.2%2026-11-15independentSource ↗
Quasar 438B48.1%2026-11-15independentSource ↗
Agnes 2.5 Pro Beta48.8%2026-11-15independentSource ↗
Ring-2.6-1T45%2026-11-15independentSource ↗
Muse Wonder Full62.6%2026-11-15independentSource ↗
Inkling 2 Small52.3%2026-11-15independentSource ↗
Nemotron 3.5 Super VL47.8%2026-11-15independentSource ↗
LFM2.5-2.6B14.4%2026-11-15independentSource ↗
Mistral Medium 3.540.2%2026-11-15independentSource ↗
Agnes 2.5 Pro Theta56%2026-11-15independentSource ↗
MiniMax M3.1 Flash53.4%2026-11-15independentSource ↗
Ling-3.0-flash-Fin42.4%2026-11-15independentSource ↗
GLM 5.359%2026-11-15independentSource ↗
gpt-oss-120b34%2026-11-15independentSource ↗
GPT-5.5 Instant52.5%2026-11-15independentSource ↗
G9v3-39A5B36.8%2026-11-15independentSource ↗
Gemini 3.1 Pro58.7%2026-11-15independentSource ↗
K2 Horizon 375B A23B42.9%2026-11-15independentSource ↗
Trinity Large Thinking40.6%2026-11-15independentSource ↗
Hy348.6%2026-11-15independentSource ↗
Qwen3.8 Max (0902)52.1%2026-11-15independentSource ↗
Motif 342.2%2026-11-15independentSource ↗
K2 Horizon 7B27.5%2026-11-15independentSource ↗
MiMo-V2.6-Pro60.9%2026-11-15independentSource ↗
Grok 4.757.4%2026-11-15independentSource ↗
Claude Haiku 5.555%2026-11-15independentSource ↗
GPT-6 Astra56.5%2026-11-15independentSource ↗
North Mini Code38.8%2026-11-15independentSource ↗
GPT-6 Luna54.6%2026-11-15independentSource ↗
GLM 5.3 Flash51.6%2026-11-15independentSource ↗
K-EXAONE 2.042%2026-11-15independentSource ↗
Solar Pro 448%2026-11-15independentSource ↗
Mistral Small 438.8%2026-11-15independentSource ↗
MiMo-V2.6-Flash51.3%2026-11-15independentSource ↗
MiniMax M347.1%2026-11-15independentSource ↗
Granite 4.2 30B37.8%2026-11-15independentSource ↗
Qwen3.5 397B A17B44.8%2026-11-15independentSource ↗
Inkling47%2026-11-15independentSource ↗
GPT-6.1 Terra55.1%2026-11-15independentSource ↗
Claude Fable 5.267.7%2026-11-15independentSource ↗
Granite 4.2 3B25.3%2026-11-15independentSource ↗
Step 558.9%2026-11-15independentSource ↗
Gemini 3.5 Flash-Lite41.3%2026-11-15independentSource ↗
GPT-6.1 Sol54.2%2026-11-15independentSource ↗
Apodex 1.145.5%2026-11-15independentSource ↗
Ling 3.1 Flash 101956.8%2026-11-15independentSource ↗
Beam 1.048.5%2026-11-15independentSource ↗
Ministral 3 3B15.3%2026-11-15independentSource ↗
Kimi K3.1 Pro60.8%2026-11-15independentSource ↗
MiniCPM5-2B26.3%2026-11-15independentSource ↗
Solar Mini 447.6%2026-11-15independentSource ↗
LongCat 2.036.3%2026-11-15independentSource ↗
Nemotron 3 Ultra40.3%2026-11-15independentSource ↗
Muse Glimmer44.9%2026-11-15independentSource ↗
DeepSeek V4.1 Flash 101755.9%2026-11-15independentSource ↗
K2 Horizon MoVA 36B A4B40%2026-11-15independentSource ↗
Claude Sonnet 5.561%2026-11-15independentSource ↗
Gemini 4 Argon61.8%2026-11-15independentSource ↗
Ling-3.0-flash-VL44.2%2026-11-15independentSource ↗
Nex-N2-Pro43.3%2026-11-15independentSource ↗
Muse Wonder60.3%2026-11-15independentSource ↗
Muse Spark Open58.4%2026-11-15independentSource ↗
Qwen3.8 27B46.6%2026-11-15independentSource ↗
K2 Horizon 3.7B22%2026-11-15independentSource ↗
Claude Opus 5.566.9%2026-11-15independentSource ↗
Command A+38.5%2026-11-15independentSource ↗
Nemotron 3.5 Lightning32.1%2026-11-15independentSource ↗

Human Baseline & Difficulty Horizon

Calibrated human reference points, specialist benchmarks, and ceiling thresholds.
Measured Human Scoren/aNot established
Baseline Protocol & Interpretation

No peer-reviewed human baseline has been published for this benchmark.

Metric & Scoring Methodology

Verification protocols, aggregation formulas, and specialized metric variants.
Primary Metric:problem pass rate (%)
Scoring Engine:unit-tests

Dataset & Compute Cost

Evaluation volume, public availability, API pricing, and local hardware requirements.
Total Dataset Size—Annotated evaluation items
Public Test SetPublicOpenly mirrored on repositories
Access GatingOpen AccessUnrestricted download
Evaluation LicenseOpen AccessDataset usage and redistribution terms
Frontier API Compute Cost:

$5 – $20 USD for full benchmark evaluation run on frontier APIs.

Recommended Local GPU Setup:

1x NVIDIA RTX 4090 (24GB) or A100 (40GB/80GB) via vLLM / SGLang

How to Run & Reproduce

Standardized evaluation protocols, CLI commands, and reproducible runner templates.
Prompt Regimezero-shot
Reasoning Modedirect
Sampling Temp0
Pass@k Budgetk = 1
Tools & SandboxPure Text
Scoring Verifierunit-tests
Option AEleutherAI LM-Evaluation-Harness (Open-Weight Models)
lm_eval --model hf --model_args pretrained=<model_path> --tasks scicode --batch_size auto
Option BOpenCompass Evaluation Framework
opencompass --datasets scicode --models <model_config>
Python APIDeterministic Inference Loop Snippet
# Standard API Evaluation Loop
from openai import OpenAI

client = OpenAI()
response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": prompt}],
    temperature=0.0,
)
Standardized Reporting Requirement:

When publishing results for SciCode, always report the exact prompt template, few-shot exemplar ordering, sampling temperature (temperature=0), maximum reasoning budget tokens, and the precise timestamped model snapshot ID.

Contamination & Memorization Analysis

Audit of pretraining exposure risks, memorization vectors, and refresh policies.
Overall Contamination Risk:MEDIUM
Refresh Cadence:

Static fixed snapshot

Test Set Exposure:

Public on web / HuggingFace