Skip to content
Weights Not Walls
Menu

GBAEval

Reported

No description yet: we only describe an eval set when we can do it correctly.

Kind
Published by a third party's leaderboard, collected by Epoch AI. The leaderboard is named and linked.
Score shown
Epoch's column "Overall score", as a percentage, higher is better. The source gives no error bars.
Results
28 in all: 7 for open-weight models, 21 for closed ones. 7 linked to a model page here.

Best open-weight result: moonshotai/Kimi-K3, 48.3%, number 14 of 28 results.

GBAEval results
#ModelWeightsSettingScoreDateSource
1claude-opus-5-5_unknownAnthropicclosedeffort unknown90.6%14 Nov 2026gbaeval.com
2gpt-6-astra_unknownOpenAIclosedeffort unknown87.1%14 Nov 2026gbaeval.com
3claude-fable-5-2_unknownAnthropicclosedeffort unknown84.8%14 Nov 2026gbaeval.com
4gpt-6.1-sol_unknownOpenAIclosedeffort unknown81.9%14 Nov 2026gbaeval.com
5claude-opus-5_unknownAnthropicclosedeffort unknown79.6%26 Jul 2026gbaeval.com
6gemini-4-argon_unknownGoogle DeepMindclosedeffort unknown78.7%14 Nov 2026gbaeval.com
7claude-fable-5_unknownAnthropicclosedeffort unknown74.5%10 Jun 2026gbaeval.com
8claude-opus-4-8_unknownAnthropicclosedeffort unknown70.9%30 May 2026gbaeval.com
9grok-4.5_unknownxAIclosedeffort unknown65.4%10 Jul 2026gbaeval.com
10claude-sonnet-5_unknownAnthropicclosedeffort unknown65.3%13 Jul 2026gbaeval.com
11gpt-5.5_unknownOpenAIclosedeffort unknown53.2%3 May 2026gbaeval.com
12gpt-5.6-sol_unknownOpenAIclosedeffort unknown52.6%15 Jul 2026gbaeval.com
13claude-sonnet-4-6_unknownAnthropicclosedeffort unknown48.8%3 May 2026gbaeval.com
14moonshotai/Kimi-K3Epoch: kimi-k3_unknown, Moonshotopeneffort unknown48.3%20 Jul 2026gbaeval.com
15gpt-5.4-2026-03-05_unknownOpenAIclosedeffort unknown45.1%12 May 2026gbaeval.com
16claude-opus-4-6_unknownAnthropicclosedeffort unknown44.1%2 May 2026gbaeval.com
17claude-opus-4-7_unknownAnthropicclosedeffort unknown43.8%2 May 2026gbaeval.com
18muse-spark-1.1Meta AIcloseddefault7.9%17 Jul 2026gbaeval.com
19gemini-3.5-flash_unknownGoogle DeepMindclosedeffort unknown6.7%20 May 2026gbaeval.com
20grok-build-0.1xAIcloseddefault2.4%8 Jun 2026gbaeval.com
21MiniMaxAI/MiniMax-M3Epoch: MiniMax-M3, MiniMaxopendefault0.9%2 Jun 2026gbaeval.com
22moonshotai/Kimi-K2.6Epoch: kimi-k2.6, Moonshotopendefault0.9%1 May 2026gbaeval.com
23moonshotai/Kimi-K2.7-CodeEpoch: kimi-k2.7-code, Moonshotopendefault0.9%21 Jun 2026gbaeval.com
24gemini-3.1-pro-previewGoogle DeepMindcloseddefault0.8%3 May 2026gbaeval.com
25qwen3.7-maxAlibabacloseddefault0.4%3 Jun 2026gbaeval.com
26zai-org/GLM-5.2Epoch: glm-5.2_unknown, Z.ai (Zhipu AI)openeffort unknown0.0%21 Jun 2026gbaeval.com
27zai-org/GLM-5.1Epoch: glm-5.1, Z.ai (Zhipu AI)opendefault0.0%not datedgbaeval.com
28MiniMaxAI/MiniMax-M2.7Epoch: MiniMax-M2.7, MiniMaxopendefault0.0%not datedgbaeval.com

Open-weight rows are shaded. Dates are when the result was run or evaluated, where the data says. Data from Epoch AI (CC-BY 4.0), fetched 19 Nov 2026: Epoch AI, ‘Capabilities & benchmarking’. Published online at epoch.ai. Retrieved from ‘https://epoch.ai/benchmarks’ [online resource]. Epoch AI. Method