Skip to main content
BenchLM
Data

Scale Labs SEAL Agentic Tool Use (Chat) (SEAL Agentic Tool Use (Chat))

We show this table for reference; we do not rank on it.

ToolComp-Chat: 485-prompt benchmark subset that needs models to chain a Python interpreter and web search calls to reach a final answer. BenchLM shows it as display-only reference data, excluded from rankings. Scale has not added a row to this board since March 2025.

Accuracy (%) on SEAL Agentic Tool Use (Chat) — November 12, 2026 snapshot

We mirror the published accuracy (%) view for SEAL Agentic Tool Use (Chat). o3-mini (high) leads the public snapshot at 63.5%, followed by Gemini 2.5 Pro Experimental (March 2025) (62.4%) and o3-mini (medium) (62.4%). We do not use these results to rank models overall.

1

o3-mini (high)

OpenAI

63.5%
2

Gemini 2.5 Pro Experimental (March 2025)

Google

62.4%
3

o3-mini (medium)

OpenAI

62.4%

35 modelsAgenticCurrentDisplay onlyUpdated November 12, 2026 snapshot

Accuracy (%) table (35 models)

Accuracy (%) results for SEAL Agentic Tool Use (Chat)
RankModel / configurationScoreParameters (B)Open / closed
1o3-mini (high)OpenAI
63.5%
Not reportedUnknown
2Gemini 2.5 Pro Experimental (March 2025)Google
62.4%
Not reportedUnknown
3o3-mini (medium)OpenAI
62.4%
Not reportedUnknown
4o1 Pro (March 2025)OpenAI
61.4%
Not reportedUnknown
5DeepSeek-R1DeepSeek
60.9%
Not reportedOpen
6o1 (December 2024)OpenAI
60.4%
Not reportedUnknown
7DeepSeek-V3 (December 2024)Deepseek
58.5%
Not reportedUnknown
8Gemini 2.0 Pro Experimental (February 2025)Google
57.9%
Not reportedUnknown
9Gemini 2.0 Flash Thinking Experimental (January 2025)Google
57.4%
Not reportedUnknown
10GPT-4o (August 2024)OpenAI
56.9%
Not reportedUnknown
11GPT-4.5 Preview (February 2025)OpenAI
56.3%
Not reportedUnknown
12Claude 3.7 Sonnet (February 2025)Anthropic
56.3%
Not reportedUnknown
13Claude 3.5 Sonnet (June 2024)Anthropic
56.1%
Not reportedUnknown
14Claude 3.7 Sonnet Thinking (February 2025)Anthropic
55.3%
Not reportedUnknown
15o1-previewOpenAI
55.1%
Not reportedClosed
16DeepSeek V3 (March 2025)Deepseek
54.3%
Not reportedUnknown
17Gemini 2.0 Flash Experimental (December 2024)Google
53.3%
Not reportedUnknown
18GPT-4 Turbo PreviewOpenAI
53.0%
Not reportedUnknown
19Gemini 1.5 Pro (August 27, 2024)Google
51.3%
Not reportedUnknown
20Gemini 2.0 Flash Lite Preview (February 2025)Google
50.3%
Not reportedUnknown
21GPT-4o (May 2024)OpenAI
49.5%
Not reportedUnknown
22Nova ProAmazon
49.2%
Not reportedClosed
23Claude 3 OpusAnthropic
48.5%
Not reportedClosed
24Gemini 2.0 FlashGoogle
48.2%
Not reportedClosed
25Nova LIteAmazon
41.7%
Not reportedUnknown
26Nova MircoAmazon
41.0%
Not reportedUnknown
27Claude 3 SonnetAnthropic
40.4%
Not reportedUnknown
28Mistral Large 2Mistral
40.4%
Not reportedClosed
29Llama 3.1 405B InstructMeta
40.1%
Not reportedUnknown
30GPT-4OpenAI
37.9%
Not reportedUnknown
31Gemini 1.5 Pro (May 2024)Google
35.5%
Not reportedUnknown
32Llama 3.1 70B InstructMeta
33.5%
Not reportedUnknown
33GPT-4o miniOpenAI
32.8%
Not reportedClosed
34Command R+Cohere
20.2%
Not reportedUnknown
35Llama 3.1 8B InstructMeta
6.1%
Not reportedUnknown

How to read this leaderboard

Compare the published configurations as complete evaluation systems. The source can combine a base model, agent scaffold, tools, budget, and inference setting in each result.

Operator receipt: 35 sourced rows are currently displayable on this page; the leading published row is o3-mini (high) at 63.5%.

Honest limit: This Scale table is display-only context, not benchmark provenance or a weighted model-only comparison. Scale has not added a row to this board since March 2025.

SEAL Agentic Tool Use (Chat) on Scale Labs

How BenchLM shows SEAL Agentic Tool Use (Chat)

BenchLM mirrors 35 published rows from Scale Labs’ public SEAL Agentic Tool Use (Chat) leaderboard, captured on November 12, 2026 snapshot.

The table is display only. It is useful context for a published agent or model configuration, but it does not enter BenchLM’s overall or category rankings. Scale has not added a row to this board since March 2025.

Snapshot

35 published rowsScale Labs sourceNo new rowsDisplay only
SEAL Agentic Tool Use (Chat) on Scale Labs

The published SEAL Agentic Tool Use (Chat) snapshot places o3-mini (high) first at 63.5%. The third row is 1.0 points behind. The broader top-10 range is 6.6 points, so many of the published results sit in a relatively narrow band.

35 models have been evaluated on SEAL Agentic Tool Use (Chat). The benchmark falls in the Agentic category. We keep external benchmark mirrors separate from the weighted global scoring system, so these results remain source-specific evidence. SEAL Agentic Tool Use (Chat) is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.

About SEAL Agentic Tool Use (Chat)

Year

2026

Tasks

ToolComp prompts, Chat subset

Format

Accuracy (%)

Difficulty

External agent and model evaluation

ToolComp prompts require several dependent tool calls, with golden answer chains and step-level labels. The Chat subset uses two common chatbot tools. The score is final-answer accuracy in percent. BenchLM mirrors 35 published rows captured on November 12, 2026 snapshot.

Scale Labs leaderboardPublic benchmark source

Freshness and provenance

Version

SEAL Agentic Tool Use (Chat) 2026

Refresh cadence

Quarterly

Staleness state

Current

Question availability

Public benchmark set

CurrentDisplay 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.

Questions

What does SEAL Agentic Tool Use (Chat) measure?

ToolComp-Chat: 485-prompt benchmark subset that needs models to chain a Python interpreter and web search calls to reach a final answer. BenchLM shows it as display-only reference data, excluded from rankings. Scale has not added a row to this board since March 2025.

Which model leads the published SEAL Agentic Tool Use (Chat) snapshot?

o3-mini (high) currently leads the published SEAL Agentic Tool Use (Chat) snapshot with 63.5% accuracy (%). BenchLM shows this benchmark for display only and does not use it in overall rankings.

How many models are evaluated on SEAL Agentic Tool Use (Chat)?

The November 12, 2026 snapshot snapshot contains 35 AI models.

Last updated: November 12, 2026 snapshot · mirrored from the public benchmark leaderboard

Know when it’s worth switching models

The model to choose, the cheaper alternative, and the release we would wait on.

Read a sample issue

Join 7,000+ subscribers.

One email each week. Unsubscribe anytime.