# MirrorCode leaderboard

external 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/external-mirror-code

MirrorCode (Epoch AI and METR). Every model the catalog carries a published MirrorCode value for, ranked by that value.

## Benchmark record

| Field | Value |
| --- | --- |
| Category | external |
| Measure | Agentic black-box reimplementation, exact-output end-to-end tests |
| Tasks | 30 reimplementation tasks (15 programs in two languages), three attempts each |
| Difficulty | Days-long software engineering without source code |
| Published by | [MirrorCode (Epoch AI)](https://epoch.ai/benchmarks/mirrorcode) |

MirrorCode, built by Epoch AI with METR, asks AI agents to reimplement entire command-line programs from scratch without their source code, matching the original's output exactly on end-to-end and held-out tests.

## MirrorCode ranking

41 models with a published MirrorCode 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 | Agentic black-box reimplementation, exact-output end-to-end tests | Evidence |
| --- | --- | --- | --- | --- |
| 1 | [Claude Sonnet 5.5](https://modelscale.dev/models/claude-sonnet-5-5) | Anthropic | 97.8 | Observed 2026-11-21 · source benchlm:models |
| 2 | [Claude Opus 5.5](https://modelscale.dev/models/claude-opus-5-5) | Anthropic | 95.6 | Observed 2026-11-21 · source benchlm:models |
| 3 | [Claude Fable 5.1](https://modelscale.dev/models/claude-fable-5-1) | Anthropic | 93.3 | Observed 2026-11-21 · source benchlm:models |
| 4 | [Claude Opus 5](https://modelscale.dev/models/claude-opus-5) | Anthropic | 87.8 | Observed 2026-11-21 · source benchlm:models |
| 5 | [Claude Haiku 5.5](https://modelscale.dev/models/claude-haiku-5-5) | Anthropic | 78.9 | Observed 2026-11-21 · source benchlm:models |
| 6 | [GPT-6.1 Terra](https://modelscale.dev/models/gpt-6-1-terra) | OpenAI | 76.7 | Observed 2026-11-21 · source benchlm:models |
| 7 | [Claude Fable 5](https://modelscale.dev/models/claude-fable-5) | Anthropic | 70.0 | Observed 2026-11-21 · source benchlm:models |
| 8 | [GPT-6.2 Astra](https://modelscale.dev/models/gpt-6-2-astra) | OpenAI | 67.8 | Observed 2026-11-21 · source benchlm:models |
| 9 | [Claude Opus 4.8](https://modelscale.dev/models/claude-opus-4-8) | Anthropic | 65.6 | Observed 2026-11-21 · source benchlm:models |
| 10 | [DeepSeek V4.1 Flash 1017](https://modelscale.dev/models/deepseek-v4-1-flash) | DeepSeek | 64.4 | Observed 2026-11-21 · source benchlm:models |
| 11 | [Claude Sonnet 5](https://modelscale.dev/models/claude-sonnet-5) | Anthropic | 63.3 | Observed 2026-11-21 · source benchlm:models |
| 12 | [GPT-6.1 Sol](https://modelscale.dev/models/gpt-6-1-sol) | OpenAI | 61.1 | Observed 2026-11-21 · source benchlm:models |
| 13 | [GPT-6 Astra](https://modelscale.dev/models/gpt-6-astra) | OpenAI | 60.0 | Observed 2026-11-21 · source benchlm:models |
| 14 | [Gemini 4 Argon](https://modelscale.dev/models/gemini-4-argon) | Google | 58.9 | Observed 2026-11-21 · source benchlm:models |
| 15 | [Grok 4.7](https://modelscale.dev/models/grok-4-7) | xAI | 54.4 | Observed 2026-11-21 · source benchlm:models |
| 15 | [Kimi K3](https://modelscale.dev/models/kimi-3) | Moonshot AI | 54.4 | Observed 2026-11-21 · source benchlm:models |
| 15 | [Muse Spark 1.3](https://modelscale.dev/models/muse-spark-1-3) | Meta | 54.4 | Observed 2026-11-21 · source benchlm:models |
| 18 | [Qwen3.8 Max](https://modelscale.dev/models/qwen3-8-max) | Alibaba | 52.2 | Observed 2026-11-21 · source benchlm:models |
| 19 | [GPT-6 Sol](https://modelscale.dev/models/gpt-6-sol) | OpenAI | 48.9 | Observed 2026-11-21 · source benchlm:models |
| 20 | [Muse Wonder](https://modelscale.dev/models/muse-wonder) | Meta | 45.6 | Observed 2026-11-21 · source benchlm:models |
| 21 | [Claude Opus 4.7](https://modelscale.dev/models/claude-opus-4-7) | Anthropic | 43.3 | Observed 2026-11-21 · source benchlm:models |
| 22 | [Gemini 3.7 Flash](https://modelscale.dev/models/gemini-3-7-flash) | Google | 41.1 | Observed 2026-11-21 · source benchlm:models |
| 22 | [GPT-5.6 Terra](https://modelscale.dev/models/gpt-5-6-terra) | OpenAI | 41.1 | Observed 2026-11-21 · source benchlm:models |
| 22 | [Grok 4.6](https://modelscale.dev/models/grok-4-6) | xAI | 41.1 | Observed 2026-11-21 · source benchlm:models |
| 25 | [Mistral Large 4](https://modelscale.dev/models/mistral-large-4) | Mistral | 40.0 | Observed 2026-11-21 · source benchlm:models |
| 26 | [Gemini 3.8 Flash](https://modelscale.dev/models/gemini-3-8-flash) | Google | 38.9 | Observed 2026-11-21 · source benchlm:models |
| 27 | [GLM-5.2](https://modelscale.dev/models/glm-5-2) | Z.AI | 36.7 | Observed 2026-11-21 · source benchlm:models |
| 28 | [Claude Opus 4.6](https://modelscale.dev/models/claude-opus-4-6) | Anthropic | 33.3 | Observed 2026-11-21 · source benchlm:models |
| 28 | [DeepSeek V4 Pro](https://modelscale.dev/models/deepseek-v4-pro) | DeepSeek | 33.3 | Observed 2026-11-21 · source benchlm:models |
| 28 | [DeepSeek V4 Pro 0813](https://modelscale.dev/models/deepseek-v4-pro-max) | DeepSeek | 33.3 | Last good 2026-11-18 · source benchlm:models |
| 31 | [GPT-5.2](https://modelscale.dev/models/gpt-5-2) | OpenAI | 32.2 | Observed 2026-11-21 · source benchlm:models |
| 31 | [Grok 4.5](https://modelscale.dev/models/grok-4-5) | xAI | 32.2 | Observed 2026-11-21 · source benchlm:models |
| 33 | [GPT-6 Luna](https://modelscale.dev/models/gpt-6-luna) | OpenAI | 26.7 | Observed 2026-11-21 · source benchlm:models |
| 34 | [Gemini 3.6 Flash](https://modelscale.dev/models/gemini-3-6-flash) | Google | 23.3 | Observed 2026-11-21 · source benchlm:models |
| 34 | [GPT-5.4](https://modelscale.dev/models/gpt-5-4) | OpenAI | 23.3 | Observed 2026-11-21 · source benchlm:models |
| 34 | [GPT-5.5](https://modelscale.dev/models/gpt-5-5) | OpenAI | 23.3 | Observed 2026-11-21 · source benchlm:models |
| 37 | [GPT-5.6 Luna](https://modelscale.dev/models/gpt-5-6-luna) | OpenAI | 21.1 | Observed 2026-11-21 · source benchlm:models |
| 38 | [GPT-5.6 Sol](https://modelscale.dev/models/gpt-5-6-sol) | OpenAI | 20.0 | Observed 2026-11-21 · source benchlm:models |
| 39 | [Gemini 3.1 Pro](https://modelscale.dev/models/gemini-3-1-pro) | Google | 8.9 | Observed 2026-11-21 · source benchlm:models |
| 40 | [GPT-5](https://modelscale.dev/models/gpt-5) | OpenAI | 4.4 | Observed 2026-11-21 · source benchlm:models |
| 40 | [GPT-5 (high)](https://modelscale.dev/models/gpt-5-high) | OpenAI | 4.4 | Last good 2026-11-18 · 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.
