All tasks / Tools
goose
Open-source general-purpose AI agent (desktop app, CLI and API) that runs on your machine and works with cloud or local LLMs.
- Agentic AI Foundation (AAIF); originally created by Block
- Local or cloud
- Free, open source
Where it ranks
- BCodingCoding agents · Local · via DeepSeek-V4.1-Flash-1017With cloud models: tier S (via Claude Opus 5.5)
Ranked by its own benchmark result where it has one; otherwise by the best model it offers.
Facts
- Vendor
- Agentic AI Foundation (AAIF); originally created by Block
- Approach
- Coding agents
- Where it runs
- Runs fully on your hardware with local models; cloud models optional
- Interface
- Desktop app, Command line, API
- Skill level
- Intermediate: some setup or configuration
- Price
- Free, open source
- Pricing notes
- Open source and free; you pay your own model provider costs or run local models (Ollama, LM Studio).
- Source code
- Open source (Apache-2.0)
- GitHub stars
- 55.4k (18 Oct 2026)
- Last verified
- 18 Oct 2026
Price from goose-docs.ai, 18 Oct 2026.
Models it gives you
- SClaude Opus 5.5Anthropic · closed
- SClaude Sonnet 5.5Anthropic · closed
- BDeepSeek-V4.1-Flash-1017DeepSeek · open weights
- BKimi K3Moonshot AI · open weights
- BQwen3.8-2.4T-A95BAlibaba (Qwen) · open weights
- CGLM-5.3Z.ai · open weights
- CQwen3.8-27BAlibaba (Qwen) · open weights
- CDeepSeek-V4-Pro-0813DeepSeek · open weights
- CMiniMax-M3MiniMax · open weights
- CLlama 4 MaverickMeta · open weights
Its own benchmark results
| Benchmark | Result | Entry on the leaderboard | As of | Source |
|---|---|---|---|---|
| Terminal-Bench 2.0 | 54.3% | Goose + Claude Opus 4.5rank 56 of 142 entries; raw accuracy 54.33 ± 2.63 (95% CI half-width); entry date 2025-11-24; not marked verified; agent org: Block; read via the site's leaderboard-read API (page is client-rendered) | 18 Oct 2026 | tbench.ai |