---
title: "Open weights against closed · Undominated.ai"
canonical: https://undominated.ai/open-vs-closed/
description: "As of 2026-10-06, 11 rated open-weight models against 58 closed, and 76 with no recorded weights status, counted apart. The best-to-best gap with its error, frontier share, and price at matched score floors."
---

# Open weights against closed · Undominated.ai

> As of 2026-10-06, 11 rated open-weight models against 58 closed, and 76 with no recorded weights status, counted apart. The best-to-best gap with its error, frontier share, and price at matched score floors.

# Open weights against closed

Rated models only, on the LMArena general board. "Open weights" is the catalogue's own flag: the model's weights are available. It is not "open source": licence and commercial-use terms still vary, and the flag says nothing about training data or code.

As of Oct 6, 2026, **11** rated open-weight models against **58** closed; **76** rated models have no recorded weights status.

The open sample is small, and 76 of the 145 rated models sit in neither column. Everything below compares the two recorded groups, not open and closed models at large. A model with no recorded status is counted apart, never as closed.

## Best against best

### Best open-weight

[Kimi K3](https://undominated.ai/models/moonshotai__kimi-k3/)

 Score 1475.9 ± 4.76 Significance rank 11 Balanced $/M $3.76

### Best closed

[Claude Opus 5.5](https://undominated.ai/models/anthropic__claude-opus-5.5/)

 Score 1511.7 ± 9.05 Significance rank 1 Balanced $/M $8.00

The best closed model scores **35.8** points above the best open-weight one. That is more than the two half-widths together (13.81), so the leaderboard separates them.

Half-widths are LMArena's published intervals. Two scores are separable when they differ by more than the two half-widths together, the same rule as the leaderboard's [significance rank](https://undominated.ai/significance/).

## On the value frontier

Of the **12** models on the leaderboard's [value frontier](https://undominated.ai/frontier/), **2** are open-weight, **2** closed, and **8** have no recorded weights status. The frontier is the whole board's, at the balanced workload, not one drawn inside each group.

## Price at matched score floors

The cheapest rated model at or above each LMArena floor, per group, balanced workload, USD per million tokens. The two picks need not score alike above the floor.

| Floor | Cheapest open-weight | Cheapest closed | Closed ÷ open | Cheapest, no recorded status |
| --- | --- | --- | --- | --- |
| ≥ 1350 | [gpt-oss-120b](https://undominated.ai/models/openai__gpt-oss-120b/) 1365.4 · $0.065 | [GPT-6 Luna](https://undominated.ai/models/openai__gpt-6-luna/) 1391.5 · $0.200 | 3.1× | [Gemma 3 27B](https://undominated.ai/models/google__gemma-3-27b-it/) 1357.8 · $0.100 |
| ≥ 1400 | [DeepSeek V4 Flash 0423](https://undominated.ai/models/deepseek__deepseek-v4-flash/) 1432.1 · $0.113 | [GPT-5.6 Luna](https://undominated.ai/models/openai__gpt-5.6-luna/) 1431.0 · $0.450 | 4.0× | [Gemma 4 26B A4B](https://undominated.ai/models/google__gemma-4-26b-a4b-it/) 1433.9 · $0.127 |
| ≥ 1450 | [GLM 5.3](https://undominated.ai/models/z-ai__glm-5.3/) 1471.4 · $1.00 | [Gemini 3.6 Flash](https://undominated.ai/models/google__gemini-3.6-flash/) 1479.5 · $1.50 | 1.5× | [MiMo-V2.6-Flash](https://undominated.ai/models/xiaomi__mimo-v2.6-flash/) 1456.4 · $0.175 |

## What the flag is

The weights status is not read from the price feed. Open is set model by model, by hand. Closed is set the same way for **33** of the 58 closed models, and for **25** it follows from the maker’s own pages, for the lines of models it publishes no weights for: Anthropic’s models, OpenAI’s except its open gpt-oss family, and Google’s Gemini line but not Gemma. The pages are quoted in data/maker-weights.json . Anthropic says so in a statement; OpenAI’s and Google’s are terms that forbid extracting the model, weaker evidence than a statement. A model whose record does not set it has no status, which is why the third group is large. The price feed names a Hugging Face repository for **61** of those 76. A named repository is evidence, not a reviewed status, so they stay unknown here rather than being guessed into either column. The [open-weights list](https://undominated.ai/open-weights/) has every flagged model, rated or not.