SciCode
Research-level scientific coding benchmark: models write code to solve subproblems drawn from real scientific workflows across natural-science domains.
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.
| Model | Score | Date | Source Type | Provenance |
|---|---|---|---|---|
| A.X-K2 | 41% | 2026-11-15 | independent | Source ↗ |
| Llama 4 Scout | 21.3% | 2026-11-15 | independent | Source ↗ |
| Qwen3.7 Plus | 46.1% | 2026-11-15 | independent | Source ↗ |
| Ling 3.1 Flash | 54.1% | 2026-11-15 | independent | Source ↗ |
| NVIDIA Nemotron 3 Nano 4B | 16.1% | 2026-11-15 | independent | Source ↗ |
| K2 Horizon 0.9B | 6.9% | 2026-11-15 | independent | Source ↗ |
| Qwen3.8-Flash-Next | 50.6% | 2026-11-15 | independent | Source ↗ |
| Ministral 3 8B | 20.7% | 2026-11-15 | independent | Source ↗ |
| Llama 4 Maverick | 31.7% | 2026-11-15 | independent | Source ↗ |
| Qwen3.8 2.4T A95B | 54.1% | 2026-11-15 | independent | Source ↗ |
| Kimi K2.7 Code | 47.8% | 2026-11-15 | independent | Source ↗ |
| DeepSeek V4-Pro | 51% | 2026-11-15 | independent | Source ↗ |
| Granite 4.2 8B | 31.5% | 2026-11-15 | independent | Source ↗ |
| Celeris-1 | 21.6% | 2026-11-15 | independent | Source ↗ |
| Ling 3.0 Tiny | 24.2% | 2026-11-15 | independent | Source ↗ |
| Solar Open2 250B | 48% | 2026-11-15 | independent | Source ↗ |
| Nemotron 3.5 Bolt 74B-A9B | 39.7% | 2026-11-15 | independent | Source ↗ |
| DeepSeek V4 Flash Vision | 49.7% | 2026-11-15 | independent | Source ↗ |
| Mercury 2.5 | 38.5% | 2026-11-15 | independent | Source ↗ |
| Gemini 3.8 Flash | 56.6% | 2026-11-15 | independent | Source ↗ |
| Inkling Small | 49.7% | 2026-11-15 | independent | Source ↗ |
| Ministral 3 14B | 23.8% | 2026-11-15 | independent | Source ↗ |
| Gemma 4 26B A4B | 40% | 2026-11-15 | independent | Source ↗ |
| Gemma 4 31B | 45.5% | 2026-11-15 | independent | Source ↗ |
| Qwen3 Coder Next | 36.2% | 2026-11-15 | independent | Source ↗ |
| Quasar 438B | 48.1% | 2026-11-15 | independent | Source ↗ |
| Agnes 2.5 Pro Beta | 48.8% | 2026-11-15 | independent | Source ↗ |
| Ring-2.6-1T | 45% | 2026-11-15 | independent | Source ↗ |
| Muse Wonder Full | 62.6% | 2026-11-15 | independent | Source ↗ |
| Inkling 2 Small | 52.3% | 2026-11-15 | independent | Source ↗ |
| Nemotron 3.5 Super VL | 47.8% | 2026-11-15 | independent | Source ↗ |
| LFM2.5-2.6B | 14.4% | 2026-11-15 | independent | Source ↗ |
| Mistral Medium 3.5 | 40.2% | 2026-11-15 | independent | Source ↗ |
| Agnes 2.5 Pro Theta | 56% | 2026-11-15 | independent | Source ↗ |
| MiniMax M3.1 Flash | 53.4% | 2026-11-15 | independent | Source ↗ |
| Ling-3.0-flash-Fin | 42.4% | 2026-11-15 | independent | Source ↗ |
| GLM 5.3 | 59% | 2026-11-15 | independent | Source ↗ |
| gpt-oss-120b | 34% | 2026-11-15 | independent | Source ↗ |
| GPT-5.5 Instant | 52.5% | 2026-11-15 | independent | Source ↗ |
| G9v3-39A5B | 36.8% | 2026-11-15 | independent | Source ↗ |
| Gemini 3.1 Pro | 58.7% | 2026-11-15 | independent | Source ↗ |
| K2 Horizon 375B A23B | 42.9% | 2026-11-15 | independent | Source ↗ |
| Trinity Large Thinking | 40.6% | 2026-11-15 | independent | Source ↗ |
| Hy3 | 48.6% | 2026-11-15 | independent | Source ↗ |
| Qwen3.8 Max (0902) | 52.1% | 2026-11-15 | independent | Source ↗ |
| Motif 3 | 42.2% | 2026-11-15 | independent | Source ↗ |
| K2 Horizon 7B | 27.5% | 2026-11-15 | independent | Source ↗ |
| MiMo-V2.6-Pro | 60.9% | 2026-11-15 | independent | Source ↗ |
| Grok 4.7 | 57.4% | 2026-11-15 | independent | Source ↗ |
| Claude Haiku 5.5 | 55% | 2026-11-15 | independent | Source ↗ |
| GPT-6 Astra | 56.5% | 2026-11-15 | independent | Source ↗ |
| North Mini Code | 38.8% | 2026-11-15 | independent | Source ↗ |
| GPT-6 Luna | 54.6% | 2026-11-15 | independent | Source ↗ |
| GLM 5.3 Flash | 51.6% | 2026-11-15 | independent | Source ↗ |
| K-EXAONE 2.0 | 42% | 2026-11-15 | independent | Source ↗ |
| Solar Pro 4 | 48% | 2026-11-15 | independent | Source ↗ |
| Mistral Small 4 | 38.8% | 2026-11-15 | independent | Source ↗ |
| MiMo-V2.6-Flash | 51.3% | 2026-11-15 | independent | Source ↗ |
| MiniMax M3 | 47.1% | 2026-11-15 | independent | Source ↗ |
| Granite 4.2 30B | 37.8% | 2026-11-15 | independent | Source ↗ |
| Qwen3.5 397B A17B | 44.8% | 2026-11-15 | independent | Source ↗ |
| Inkling | 47% | 2026-11-15 | independent | Source ↗ |
| GPT-6.1 Terra | 55.1% | 2026-11-15 | independent | Source ↗ |
| Claude Fable 5.2 | 67.7% | 2026-11-15 | independent | Source ↗ |
| Granite 4.2 3B | 25.3% | 2026-11-15 | independent | Source ↗ |
| Step 5 | 58.9% | 2026-11-15 | independent | Source ↗ |
| Gemini 3.5 Flash-Lite | 41.3% | 2026-11-15 | independent | Source ↗ |
| GPT-6.1 Sol | 54.2% | 2026-11-15 | independent | Source ↗ |
| Apodex 1.1 | 45.5% | 2026-11-15 | independent | Source ↗ |
| Ling 3.1 Flash 1019 | 56.8% | 2026-11-15 | independent | Source ↗ |
| Beam 1.0 | 48.5% | 2026-11-15 | independent | Source ↗ |
| Ministral 3 3B | 15.3% | 2026-11-15 | independent | Source ↗ |
| Kimi K3.1 Pro | 60.8% | 2026-11-15 | independent | Source ↗ |
| MiniCPM5-2B | 26.3% | 2026-11-15 | independent | Source ↗ |
| Solar Mini 4 | 47.6% | 2026-11-15 | independent | Source ↗ |
| LongCat 2.0 | 36.3% | 2026-11-15 | independent | Source ↗ |
| Nemotron 3 Ultra | 40.3% | 2026-11-15 | independent | Source ↗ |
| Muse Glimmer | 44.9% | 2026-11-15 | independent | Source ↗ |
| DeepSeek V4.1 Flash 1017 | 55.9% | 2026-11-15 | independent | Source ↗ |
| K2 Horizon MoVA 36B A4B | 40% | 2026-11-15 | independent | Source ↗ |
| Claude Sonnet 5.5 | 61% | 2026-11-15 | independent | Source ↗ |
| Gemini 4 Argon | 61.8% | 2026-11-15 | independent | Source ↗ |
| Ling-3.0-flash-VL | 44.2% | 2026-11-15 | independent | Source ↗ |
| Nex-N2-Pro | 43.3% | 2026-11-15 | independent | Source ↗ |
| Muse Wonder | 60.3% | 2026-11-15 | independent | Source ↗ |
| Muse Spark Open | 58.4% | 2026-11-15 | independent | Source ↗ |
| Qwen3.8 27B | 46.6% | 2026-11-15 | independent | Source ↗ |
| K2 Horizon 3.7B | 22% | 2026-11-15 | independent | Source ↗ |
| Claude Opus 5.5 | 66.9% | 2026-11-15 | independent | Source ↗ |
| Command A+ | 38.5% | 2026-11-15 | independent | Source ↗ |
| Nemotron 3.5 Lightning | 32.1% | 2026-11-15 | independent | Source ↗ |
Human Baseline & Difficulty Horizon
Calibrated human reference points, specialist benchmarks, and ceiling thresholds.No peer-reviewed human baseline has been published for this benchmark.
Metric & Scoring Methodology
Verification protocols, aggregation formulas, and specialized metric variants.problem pass rate (%)Dataset & Compute Cost
Evaluation volume, public availability, API pricing, and local hardware requirements.$5 – $20 USD for full benchmark evaluation run on frontier APIs.
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.lm_eval --model hf --model_args pretrained=<model_path> --tasks scicode --batch_size autoopencompass --datasets scicode --models <model_config># 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,
)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.Static fixed snapshot
Public on web / HuggingFace