# SciCode leaderboard

Coding benchmark

> Every figure below is reproduced as its upstream source published it: nothing is modelled, estimated, interpolated or converted. `Unavailable` means no source published the value — it is never a zero. Each value carries its evidence state and the date it was observed.

Page: https://modelscale.dev/leaderboards/benchmarks/coding-sci-code

Scientific Code Benchmark. Every model the catalog carries a published SciCode value for, ranked by that value.

## Benchmark record

| Field | Value |
| --- | --- |
| Category | Coding |
| Measure | — |
| Tasks | — |
| Difficulty | — |
| Published by | [BenchLM](https://benchlm.ai/benchmarks/scicode) |

SciCode evaluates language models on generating code for realistic scientific research problems across 16 subfields of physics, math, chemistry, biology, and material science. Problems decompose into 338 subproblems requiring domain knowledge recall, scientific reasoning, and precise code synthesis. Based on real scripts from published research.

## SciCode ranking

35 models with a published SciCode value, ordered by that value, highest first. Models the source has not scored on this benchmark are not listed — they are unmeasured, not last.

| Rank | Model | Provider | SciCode | Evidence |
| --- | --- | --- | --- | --- |
| 1 | [Sakana Fugu](https://modelscale.dev/models/sakana-fugu) | Sakana AI | 60.1 | Observed 2026-11-21 · source benchlm:models |
| 2 | [Step 5 Preview](https://modelscale.dev/models/step-5-preview) | StepFun | 58.9 | Observed 2026-11-21 · source benchlm:models |
| 3 | [Sakana Fugu-Ultra](https://modelscale.dev/models/sakana-fugu-ultra) | Sakana AI | 58.7 | Observed 2026-11-21 · source benchlm:models |
| 4 | [Qwen3.7 Max](https://modelscale.dev/models/qwen3-7-max) | Alibaba | 53.5 | Observed 2026-11-21 · source benchlm:models |
| 5 | [Gemini 3.5 Flash](https://modelscale.dev/models/gemini-3-5-flash) | Google | 53.1 | Observed 2026-11-21 · source benchlm:models |
| 6 | [Kimi K2.6](https://modelscale.dev/models/kimi-2-6) | Moonshot AI | 52.2 | Observed 2026-11-21 · source benchlm:models |
| 7 | [Qwen3.7 Plus](https://modelscale.dev/models/qwen3-7-plus) | Alibaba | 51.3 | Observed 2026-11-21 · source benchlm:models |
| 8 | [Beam 1.0 501B-A23B](https://modelscale.dev/models/reflection-beam) | Reflection AI | 49.2 | Observed 2026-11-21 · source benchlm:models |
| 9 | [Inkling-Small](https://modelscale.dev/models/inkling-small) | Thinking Machines Lab | 48.7 | Observed 2026-11-21 · source benchlm:models |
| 9 | [Kimi K2.5](https://modelscale.dev/models/kimi-k2-5) | Moonshot AI | 48.7 | Observed 2026-11-21 · source benchlm:models |
| 11 | [Beam 1.0 501B-A23B FP8](https://modelscale.dev/models/reflection-beam-fp8) | Reflection AI | 48.3 | Observed 2026-11-21 · source benchlm:models |
| 12 | [Grok 4.3](https://modelscale.dev/models/grok-4-3) | xAI | 47.3 | Observed 2026-11-21 · source benchlm:models |
| 13 | [Qwen 3.6 Max (preview)](https://modelscale.dev/models/qwen3-6-max-preview) | Alibaba | 47 | Observed 2026-11-21 · source benchlm:models |
| 14 | [Beam 1.0 501B-A23B NVFP4](https://modelscale.dev/models/reflection-beam-nvfp4) | Reflection AI | 46.9 | Observed 2026-11-21 · source benchlm:models |
| 15 | [Nemotron 3 Ultra](https://modelscale.dev/models/nemotron-3-ultra-500b) | NVIDIA | 44.6 | Observed 2026-11-21 · source benchlm:models |
| 16 | [Sage-A1.5](https://modelscale.dev/models/sage-a1-5) | InternScience | 44.3 | Observed 2026-11-21 · source benchlm:models |
| 17 | [Muse Glimmer 30B](https://modelscale.dev/models/muse-glimmer-30b) | Meta | 43.6 | Observed 2026-11-21 · source benchlm:models |
| 18 | [Ling 3.0 Flash](https://modelscale.dev/models/ling-3-0-flash) | InclusionAI | 41.24 | Observed 2026-11-21 · source benchlm:models |
| 19 | [Hy3 Preview](https://modelscale.dev/models/hy3-preview) | Tencent | 41.2 | Observed 2026-11-21 · source benchlm:models |
| 20 | [A.X K2](https://modelscale.dev/models/a-x-k2) | SK Telecom | 41 | Observed 2026-11-21 · source benchlm:models |
| 21 | [Ling 3.0 Flash FP8](https://modelscale.dev/models/ling-3-0-flash-fp8) | InclusionAI | 40.37 | Observed 2026-11-21 · source benchlm:models |
| 22 | [Ripple-65B Preview](https://modelscale.dev/models/ripple-65b-preview) | Ricursive Intelligence | 39.4 | Observed 2026-11-21 · source benchlm:models |
| 23 | [Granite 4.2 30B](https://modelscale.dev/models/granite-4-2-30b) | IBM | 38.76 | Observed 2026-11-21 · source benchlm:models |
| 24 | [Mercury 2.5](https://modelscale.dev/models/mercury-2-5) | Inception | 38 | Observed 2026-11-21 · source benchlm:models |
| 25 | [K-EXAONE 2.0](https://modelscale.dev/models/k-exaone-2-0) | LG AI Research | 37.4 | Observed 2026-11-21 · source benchlm:models |
| 26 | [Spark-X2.6-4B](https://modelscale.dev/models/spark-x2-6-4b) | SparkLLM | 36.8 | Observed 2026-11-21 · source benchlm:models |
| 27 | [Granite 4.2 8B](https://modelscale.dev/models/granite-4-2-8b) | IBM | 36.09 | Observed 2026-11-21 · source benchlm:models |
| 28 | [LFM2.5-24B-A2B](https://modelscale.dev/models/lfm2-5-24b-a2b) | LiquidAI | 32.71 | Observed 2026-11-21 · source benchlm:models |
| 29 | [Nemotron 3 Nano Omni 30B A3B](https://modelscale.dev/models/nemotron-3-nano-omni-30b-a3b) | NVIDIA | 32 | Observed 2026-11-21 · source benchlm:models |
| 30 | [Nemotron 3.5 Lightning 30B A3B NVFP4](https://modelscale.dev/models/nemotron-3-5-lightning-30b-a3b-nvfp4) | NVIDIA | 31.38 | Observed 2026-11-21 · source benchlm:models |
| 31 | [Agents-A1-4B](https://modelscale.dev/models/agents-a1-4b) | InternScience | 29.6 | Observed 2026-11-21 · source benchlm:models |
| 32 | [LFM3-1.7B](https://modelscale.dev/models/lfm3-1-7b) | LiquidAI | 28.47 | Observed 2026-11-21 · source benchlm:models |
| 33 | [Ling 2.6 Flash](https://modelscale.dev/models/ling-2-6-flash) | InclusionAI | 27 | Observed 2026-11-21 · source benchlm:models |
| 34 | [MiniCPM5-2B](https://modelscale.dev/models/minicpm5-2b) | OpenBMB | 26.3 | Observed 2026-11-21 · source benchlm:models |
| 35 | [Granite 4.2 3B](https://modelscale.dev/models/granite-4-2-3b) | IBM | 24.11 | Observed 2026-11-21 · source benchlm:models |

Rows are ordered by the value the source published, highest first. The source does not state whether a higher value is the better result, so this document does not either: for a benchmark that measures a rate of failure, read the table from the bottom.
