Skip to content
← Models
Open weightsTeam-ACE/10 trustMetrics-checked

ToolACE-8B

8B function-calling model trained on self-evolved tool data; rivaled GPT-4 on the BFCL leaderboard.

Visit official page →

About

ToolACE-8B is a function-calling language model from Team-ACE, introduced in the September 2024 paper 'ToolACE: Winning the Points of LLM Function Calling' (arXiv:2409.00920). Rather than hand-annotating tool data, the team built an automatic agentic pipeline with a self-evolution synthesis process that curated a pool of 26,507 diverse APIs, generated multi-agent dialogues guided by a formalized thinking process, and verified data accuracy with a dual-layer rule-based plus model-based check system. The resulting 8B model achieved state-of-the-art results on the Berkeley Function-Calling Leaderboard, rivaling the latest GPT-4 models at a fraction of the size. The model and a subset of the data were released publicly on Hugging Face (Team-ACE/ToolACE-8B).

Limitations: ToolACE is a function-calling specialist, not a general-purpose chat model — general reasoning and knowledge tasks are not its strength. The underlying base model and exact weight license should be confirmed on the Hugging Face model card before commercial deployment. Newer agentic models have since advanced the state of the art, so treat its headline results as 2024-era.

Specs

Context window

Pricing

Free quota

API access

Weights

Open weights

Released

September 2, 2024

Benchmarks

Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.

Use cases

function-callingagentstool-use

Trust score — 6/10

/10 trustMetrics-checked

Scored 6/10 (metrics tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 68; Trending score 64/100 from researcher assessment.

Sources