Benchmark profile

General AI Assistants (GAIA)

GAIA evaluates AI models on real-world tasks that are conceptually simple for humans but require multi-step reasoning, web browsing, tool use, and multimodal understanding for AI. Tasks span three difficulty levels and test practical assistant capabilities rather than academic knowledge.

Data verified July 23, 2026

How BenchLM shows GAIA right now

BenchLM is tracking GAIA in the local dataset, but exact-source verification records for these rows are still being attached. To avoid a blank benchmark page, BenchLM shows the current tracked rows below as a display-only reference table.

These tracked rows are useful for inspection and spot-checking, but until exact-source attachments are completed they should not be treated as fully verified public benchmark rows.

27 tracked modelsLocal tracked rowsAwaiting exact-source attachmentsDisplay only

Tracked score on GAIA — July 23, 2026

BenchLM mirrors the published tracked score view for GAIA. Claude Mythos 5 leads the public snapshot at 52.3% , followed by Claude Fable 5 (52.3%) and GPT-5.4 Pro (50.5%). BenchLM does not use these results to rank models overall.

1

Claude Mythos 5

Anthropic

claude-mythos-5

52.3%

Overall —

2

Claude Fable 5

Anthropic

claude-fable-5

52.3%

Overall —

3

GPT-5.4 Pro

OpenAI

gpt-5-4-pro

50.5%

Overall —

27 modelsAgenticRefreshingDisplay onlyUpdated July 23, 2026

Tracked score table (27 models)

Score

1

Claude Mythos 5Anthropic

52.3%

2

Claude Fable 5Anthropic

52.3%

3

GPT-5.4 ProOpenAI

50.5%

4

GPT-5.4OpenAI

48.2%

5

Claude Opus 4.6Anthropic

47.8%

6

Gemini 3.1 ProGoogle

46.1%

7

Claude Sonnet 4.6Anthropic

45.5%

8

GPT-5 miniOpenAI

44.8%

9

GPT-5 (high)OpenAI

42.1%

10

GPT-5.2OpenAI

40.3%

11

Grok 4.1xAI

39.7%

12

Gemini 3 ProGoogle

38.5%

13

Kimi K2.5 (Reasoning)Moonshot AI

38.1%

14

Qwen3.6 PlusAlibaba

37.4%

15

o4-mini (high)OpenAI

36.8%

16

Kimi K2.5Moonshot AI

36.5%

17

Qwen3.5 397B (Reasoning)Alibaba

35.6%

18

Gemini 3 FlashGoogle

35.2%

19

DeepSeek V3.2 (Thinking)DeepSeek

34.8%

20

Llama 4 BehemothMeta

34.2%

21

GLM-5 (Reasoning)Z.AI

33.8%

22

Qwen3.5 397BAlibaba

33.2%

23

DeepSeek V3.2DeepSeek

32.4%

24

GLM-5Z.AI

31.5%

25

Mistral Large 3Mistral

30.8%

26

Step 3.5 FlashStepFun

29.4%

27

Llama 4 MaverickMeta

28.6%

The published GAIA snapshot places Claude Mythos 5 first at 52.3%. The third row is 1.8 points behind. The broader top-10 range is 12.0 points, so the table still separates the published systems.

27 models have been evaluated on GAIA. The benchmark falls in the Agentic category. This category carries a 22% weight in BenchLM.ai's overall scoring system. GAIA is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.

About GAIA

Year

2024

Tasks

466

Public benchmark source

BenchLM freshness & provenance

Version

GAIA 2024

Refresh cadence

Annual

Staleness state

Refreshing

Question availability

Public benchmark set

RefreshingDisplay only

BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.

FAQ

What does GAIA measure?

GAIA evaluates AI models on real-world tasks that are conceptually simple for humans but require multi-step reasoning, web browsing, tool use, and multimodal understanding for AI. Tasks span three difficulty levels and test practical assistant capabilities rather than academic knowledge.

Which model leads the published GAIA snapshot?

Claude Mythos 5 currently leads the published GAIA snapshot with 52.3% tracked score. BenchLM shows this benchmark for display only and does not use it in overall rankings.

How many models are evaluated on GAIA?

27 AI models are included in BenchLM's mirrored GAIA snapshot, based on the public leaderboard captured on July 23, 2026.

Last updated: July 23, 2026 · mirrored from the public benchmark leaderboard

Choose a model with this week’s evidence

Join 2,000+ readers for ranking moves, pricing changes, and the claims that still need proof.

One email each week. Unsubscribe anytime.