# GDPval Extended — Mercor evaluation (GDPval Extended (Mercor))

> Producing workplace deliverables from reference files. This table shows Mercor-run configurations for reference and is excluded from model rankings.

Canonical page: https://benchlm.ai/benchmarks/mercorgdpvalextended

- Category: [Agentic](/agentic)
- Last updated: October 15, 2026 capture

Evaluation results published by [Mercor](https://www.mercor.com/apex/mercor-extended/ots-held-out-gdpval-leaderboard/). Published evaluation aggregates only; task contents and dataset license grants are not included.

We captured Mercor's GDPval Extended results on October 15, 2026. The table preserves 42 published configurations, their source identifiers, reasoning settings, and disclosed mean score values. These are results from Mercor's evaluation setup; we did not rerun them.

The headline uses the 132 Mercor-built tasks. The paired original-set values refer to the separately listed 213-task public benchmark. Missing baselines remain unreported. We keep the extension and original set separate from their upstream benchmark-owner and provider-run tables.

The headline is the published mean score, which can include partial credit. It must not be read as a strict task-completion rate.

The row labels retain published error margins and sample counts when supplied. Missing margins and counts remain unreported; a sample count is not used to reconstruct a task denominator. Except where the dedicated source defines an interval, the confidence level is unreported. Equal displayed scores and overlapping margins do not establish statistical ties. Capture time does not establish individual evaluation dates.

The general methodology lists llm-as-a-judge (lmaaj) and a judge of Gemini 3 Flash Preview. Its run-count label is k = 4. Those notes describe the original public benchmark; extension-specific changes are not established by that reference.

- [Mercor GDPval Extended results](https://www.mercor.com/apex/mercor-extended/ots-held-out-gdpval-leaderboard/)
- [Mercor APEX benchmark index](https://www.mercor.com/apex/)
- [Mercor methodology](https://www.mercor.com/apex/methodology/)
- [Paired Mercor original-set results](https://www.mercor.com/apex/oss-benchmarks/oss-gdpval-leaderboard/)
- [Source paper](https://arxiv.org/abs/2510.04374)
- [Upstream dataset](https://huggingface.co/datasets/openai/gdpval)

## About GDPval Extended (Mercor)

- Tasks: 132 Mercor tasks; 213 tasks in the separately reported original set
- Format: Mean Score
- Difficulty: Source-specific evaluation
- Paper: [GDPval Extended evaluation results](https://www.mercor.com/apex/mercor-extended/ots-held-out-gdpval-leaderboard/)

The selected headline is Mean Score. Source model identifiers, effort settings, published margins, and reported sample counts remain attached to the source configuration. Public-set results and Mercor task extensions occupy separate tables.

GDPval Extended (Mercor) is currently displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.

## Leaderboard (42 configurations)

| Rank | Model | Configuration | Creator | Score |
|------|-------|---------------|---------|-------|
| 1 | [Opus 5.5](https://www.mercor.com/apex/mercor-extended/ots-held-out-gdpval-leaderboard/) | Mercor extended-set run · max reasoning · Published margin ±3 · Original set 88.3% ±1.5 · 516 reported samples | Anthropic | 75.80% |
| 2 | [Sonnet 5.5](https://www.mercor.com/apex/mercor-extended/ots-held-out-gdpval-leaderboard/) | Mercor extended-set run · max reasoning · Published margin ±3.3 · Original set 86.5% ±1.7 · 505 reported samples | Anthropic | 70.90% |
| 3 | [Fable 5.1](https://www.mercor.com/apex/mercor-extended/ots-held-out-gdpval-leaderboard/) | Mercor extended-set run · max reasoning · Published margin ±3.1 · Original set 86.6% ±1.7 · 528 reported samples | Anthropic | 68.70% |
| 4 | [Haiku 5.5](https://www.mercor.com/apex/mercor-extended/ots-held-out-gdpval-leaderboard/) | Mercor extended-set run · max reasoning · Published margin ±3.1 · Original set 85.5% ±1.6 · 506 reported samples | Anthropic | 66.00% |
| 5 | [Fable 5](https://www.mercor.com/apex/mercor-extended/ots-held-out-gdpval-leaderboard/) | Mercor extended-set run · max reasoning · Published margin ±2.9 · Original set 78.6% ±5.1 · 521 reported samples | Anthropic | 65.30% |
| 6 | [GPT-6 Astra](https://www.mercor.com/apex/mercor-extended/ots-held-out-gdpval-leaderboard/) | Mercor extended-set run · max reasoning · Published margin ±3.1 · Original set 86.7% ±1.7 · 528 reported samples | OpenAI | 64.10% |
| 7 | [Opus 5](https://www.mercor.com/apex/mercor-extended/ots-held-out-gdpval-leaderboard/) | Mercor extended-set run · max reasoning · Published margin ±3.3 · Original set 85.6% ±1.9 · 527 reported samples | Anthropic | 62.90% |
| 8 | [Grok 4.6](https://www.mercor.com/apex/mercor-extended/ots-held-out-gdpval-leaderboard/) | Mercor extended-set run · xhigh reasoning · Published margin ±2.9 · Original set 86% ±1.6 · 526 reported samples | xAI | 61.80% |
| 9 | [Fable 5.1](https://www.mercor.com/apex/mercor-extended/ots-held-out-gdpval-leaderboard/) | Mercor extended-set run · high reasoning · Published margin ±4 · Original set 81.4% ±2.7 · 528 reported samples | Anthropic | 61.50% |
| 10 | [Gemini 3.8 Flash](https://www.mercor.com/apex/mercor-extended/ots-held-out-gdpval-leaderboard/) | Mercor extended-set run · high reasoning · Published margin ±3.1 · Original set 83.5% ±1.6 · 528 reported samples | Google | 60.40% |
| 11 | [Opus 4.8](https://www.mercor.com/apex/mercor-extended/ots-held-out-gdpval-leaderboard/) | Mercor extended-set run · max reasoning · Published margin ±2.9 · Original set 84.5% ±1.7 · 528 reported samples | Anthropic | 58.90% |
| 12 | [GPT-6 Sol](https://www.mercor.com/apex/mercor-extended/ots-held-out-gdpval-leaderboard/) | Mercor extended-set run · max reasoning · Published margin ±3.1 · Original set 83.5% ±1.9 · 528 reported samples | OpenAI | 58.70% |
| 13 | [DeepSeek-V4.1-Flash](https://www.mercor.com/apex/mercor-extended/ots-held-out-gdpval-leaderboard/) | Mercor extended-set run · max reasoning · Published margin ±3 · Original set 86.7% ±1.5 · 478 reported samples | DeepSeek | 58.60% |
| 14 | [Gemini 3.7 Flash](https://www.mercor.com/apex/mercor-extended/ots-held-out-gdpval-leaderboard/) | Mercor extended-set run · high reasoning · Published margin ±2.9 · Original set 85.2% ±1.7 · 528 reported samples | Google | 57.80% |
| 15 | [Muse Spark 1.2](https://www.mercor.com/apex/mercor-extended/ots-held-out-gdpval-leaderboard/) | Mercor extended-set run · xhigh reasoning · Published margin ±3.2 · Original set 83.8% ±2 · 524 reported samples | Meta | 57.50% |
| 16 | [Kimi K3](https://www.mercor.com/apex/mercor-extended/ots-held-out-gdpval-leaderboard/) | Mercor extended-set run · max reasoning · Published margin ±3.2 · Original set 78.8% ±5.1 · 526 reported samples | Kimi | 56.40% |
| 17 | [Muse Spark 1.3](https://www.mercor.com/apex/mercor-extended/ots-held-out-gdpval-leaderboard/) | Mercor extended-set run · max reasoning · Published margin ±3.5 · Original set not reported · 527 reported samples | Meta | 55.90% |
| 18 | [GPT-5.5](https://www.mercor.com/apex/mercor-extended/ots-held-out-gdpval-leaderboard/) | Mercor extended-set run · xhigh reasoning · Published margin ±2.6 · Original set 79.9% ±5.2 · 528 reported samples | OpenAI | 54.50% |
| 19 | [Sonnet 5](https://www.mercor.com/apex/mercor-extended/ots-held-out-gdpval-leaderboard/) | Mercor extended-set run · max reasoning · Published margin ±3 · Original set 77.6% ±5.2 · 528 reported samples | Anthropic | 54.20% |
| 20 | [Opus 4.7](https://www.mercor.com/apex/mercor-extended/ots-held-out-gdpval-leaderboard/) | Mercor extended-set run · max reasoning · Published margin ±2.7 · Original set 79.8% ±5.1 · 528 reported samples | Anthropic | 53.80% |
| 21 | [Opus 4.6](https://www.mercor.com/apex/mercor-extended/ots-held-out-gdpval-leaderboard/) | Mercor extended-set run · max reasoning · Published margin ±2.8 · Original set not reported · 528 reported samples | Anthropic | 53.00% |
| 22 | [Muse Spark 1.1](https://www.mercor.com/apex/mercor-extended/ots-held-out-gdpval-leaderboard/) | Mercor extended-set run · xhigh reasoning · Published margin ±3.1 · Original set 82.4% ±2 · 522 reported samples | Meta | 52.40% |
| 23 | [Gemini 3.6 Flash](https://www.mercor.com/apex/mercor-extended/ots-held-out-gdpval-leaderboard/) | Mercor extended-set run · high reasoning · Published margin ±2.9 · Original set 83.8% ±1.7 · 527 reported samples | Google | 52.20% |
| 24 | [GPT-5.6 Sol (Pro)](https://www.mercor.com/apex/mercor-extended/ots-held-out-gdpval-leaderboard/) | Mercor extended-set run · max reasoning · Published margin ±3.1 · Original set 75.6% ±5.3 · 528 reported samples | OpenAI | 52.00% |
| 25 | [Qwen3.8-Max](https://www.mercor.com/apex/mercor-extended/ots-held-out-gdpval-leaderboard/) | Mercor extended-set run · xhigh reasoning · Published margin ±3.5 · Original set 83.3% ±2 · 528 reported samples | Alibaba | 51.10% |
| 26 | [Sonnet 4.6](https://www.mercor.com/apex/mercor-extended/ots-held-out-gdpval-leaderboard/) | Mercor extended-set run · high reasoning · Published margin ±2.8 · Original set 79% ±5.2 · 528 reported samples | Anthropic | 51.00% |
| 27 | [Grok 4.5](https://www.mercor.com/apex/mercor-extended/ots-held-out-gdpval-leaderboard/) | Mercor extended-set run · high reasoning · Published margin ±3.3 · Original set 78.8% ±5.2 · 528 reported samples | xAI | 51.00% |
| 28 | [GPT-6 Luna](https://www.mercor.com/apex/mercor-extended/ots-held-out-gdpval-leaderboard/) | Mercor extended-set run · max reasoning · Published margin ±2.9 · Original set 80.1% ±2.2 · 527 reported samples | OpenAI | 48.90% |
| 29 | [GPT-5.4](https://www.mercor.com/apex/mercor-extended/ots-held-out-gdpval-leaderboard/) | Mercor extended-set run · xhigh reasoning · Published margin ±2.7 · Original set 60% ±5.7 · 528 reported samples | OpenAI | 48.00% |
| 30 | [GLM-5.2](https://www.mercor.com/apex/mercor-extended/ots-held-out-gdpval-leaderboard/) | Mercor extended-set run · max reasoning · Published margin ±3 · Original set 78.2% ±5.2 · 529 reported samples | Zhipu | 47.30% |
| 31 | [GPT-5.6 Terra](https://www.mercor.com/apex/mercor-extended/ots-held-out-gdpval-leaderboard/) | Mercor extended-set run · max reasoning · Published margin ±3 · Original set 77.1% ±5.2 · 528 reported samples | OpenAI | 47.20% |
| 32 | [GPT-5.6 Luna](https://www.mercor.com/apex/mercor-extended/ots-held-out-gdpval-leaderboard/) | Mercor extended-set run · max reasoning · Published margin ±3 · Original set 72.5% ±5.5 · 528 reported samples | OpenAI | 44.90% |
| 33 | [DeepSeek-V4-Pro](https://www.mercor.com/apex/mercor-extended/ots-held-out-gdpval-leaderboard/) | Mercor extended-set run · max reasoning · Published margin ±3.1 · Original set 74.3% ±5.1 · 528 reported samples | DeepSeek | 42.00% |
| 34 | [Gemini 3.1 Pro](https://www.mercor.com/apex/mercor-extended/ots-held-out-gdpval-leaderboard/) | Mercor extended-set run · high reasoning · Published margin ±2.7 · Original set 68.5% ±5.5 · 527 reported samples | Google | 41.90% |
| 35 | [Kimi K2.7 Code](https://www.mercor.com/apex/mercor-extended/ots-held-out-gdpval-leaderboard/) | Mercor extended-set run · high reasoning · Published margin ±2.8 · Original set 70.1% ±5.4 · 528 reported samples | Kimi | 41.00% |
| 36 | [Inkling](https://www.mercor.com/apex/mercor-extended/ots-held-out-gdpval-leaderboard/) | Mercor extended-set run · high reasoning · Published margin ±2.8 · Original set 65.4% ±5.6 · 528 reported samples | Thinking Machines | 39.50% |
| 37 | [MiniMax-M3](https://www.mercor.com/apex/mercor-extended/ots-held-out-gdpval-leaderboard/) | Mercor extended-set run · high reasoning · Published margin ±2.9 · Original set 74.8% ±5.3 · 528 reported samples | MiniMax | 38.80% |
| 38 | [Nemotron 3 Ultra](https://www.mercor.com/apex/mercor-extended/ots-held-out-gdpval-leaderboard/) | Mercor extended-set run · high reasoning · Published margin ±2.6 · Original set 63.5% ±3.1 · 528 reported samples | NVIDIA | 38.20% |
| 39 | [Qwen3.5](https://www.mercor.com/apex/mercor-extended/ots-held-out-gdpval-leaderboard/) | Mercor extended-set run · Published margin ±2.9 · Original set 71% ±2.2 · 528 reported samples | Alibaba | 35.40% |
| 40 | [MiniMax-M2.7](https://www.mercor.com/apex/mercor-extended/ots-held-out-gdpval-leaderboard/) | Mercor extended-set run · high reasoning · Published margin ±2.5 · Original set 68.3% ±5.3 · 528 reported samples | MiniMax | 32.40% |
| 41 | [Gemini 3.5 Flash](https://www.mercor.com/apex/mercor-extended/ots-held-out-gdpval-leaderboard/) | Mercor extended-set run · high reasoning · Published margin ±4.4 · Original set 72.4% ±5.2 · 527 reported samples | Google | 31.60% |
| 42 | [Kimi K2 (Thinking)](https://www.mercor.com/apex/mercor-extended/ots-held-out-gdpval-leaderboard/) | Mercor extended-set run · high reasoning · Published margin ±2.5 · Original set 54.1% ±5.6 · 528 reported samples | Kimi | 26.00% |

## FAQ

### What does GDPval Extended (Mercor) measure?

Producing workplace deliverables from reference files. This table shows Mercor-run configurations for reference and is excluded from model rankings.

### Which model leads the published GDPval Extended (Mercor) snapshot?

Opus 5.5 currently leads the published GDPval Extended (Mercor) snapshot with a score of 75.80%.

### How many models are evaluated on GDPval Extended (Mercor)?

The October 15, 2026 capture contains 42 source configurations.

### Does GDPval Extended (Mercor) affect BenchLM's overall score?

Not directly. GDPval Extended (Mercor) is still displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.