# Best LLM for Goose

> Updated Oct 22, 2026 · evidence through Oct 22, 2026. Synthesized from cited community evidence and open benchmarks, not vendor marketing. Canonical page: https://modelmatchbook.com/best/agents-goose

Claude Opus 5.5 ranks first for Goose as of October 22, 2026. Leads both Terminal-Bench 4.0 and LMArena's agentic arena, scoring 64.85% on reproducible terminal tasks and ranking #1 with a 20.9 score on human preference for tool-use. Claude Sonnet 5.5 is the next-ranked alternative. Strong second-place showing on Terminal-Bench 4.0 at 61.82%, though its LMArena agentic rank drops to #5, suggesting solid terminal skills with more variable multi-step performance. [1,2,3,4]

## 1. Claude Opus 5.5
https://modelmatchbook.com/models/anthropic/claude-opus-5.5-20260921

Leads both Terminal-Bench 4.0 and LMArena's agentic arena, scoring 64.85% on reproducible terminal tasks and ranking #1 with a 20.9 score on human preference for tool-use.

**Best when:** Use for complex repository automation requiring precise multi-step terminal execution, as it tops Terminal-Bench 4.0 at 64.85% with Claude Code. [1]

**More tips:**
- Deploy when human preference for agentic tool-use matters most, given its #1 rank (score 20.9) on LMArena's agentic arena. [2]

## 2. Claude Sonnet 5.5
https://modelmatchbook.com/models/anthropic/claude-sonnet-5.5-20260928

Strong second-place showing on Terminal-Bench 4.0 at 61.82%, though its LMArena agentic rank drops to #5, suggesting solid terminal skills with more variable multi-step performance.

**Best when:** Choose for terminal-heavy workflows where near-top performance suffices, scoring 61.82% on Terminal-Bench 4.0 with Claude Code. [3]

**Watch out for:**
- Expect less consistent human preference on open-ended agentic tasks, ranking #5 (score 12.7) versus the leader's 20.9 on LMArena's agentic arena. [4]

## 3. GPT-6.1 Sol
https://modelmatchbook.com/models/openai/gpt-6.1-sol-20260929

Scores 58.18% on Terminal-Bench 4.0 but ranks lower on LMArena agentic preference at #4, showing capable terminal execution with less favored multi-step behavior.

**Best when:** Use for terminal automation where Codex integration is preferred, achieving 58.18% on Terminal-Bench 4.0. [7]

## 4. GPT-6 Astra
https://modelmatchbook.com/models/openai/gpt-6-astra-20260903

Matches Terminal-Bench scores with GPT-6.1 Sol at 58.18% but achieves higher LMArena agentic preference at #2, indicating stronger relative appeal for multi-step tool use.

**Best when:** Prioritize when LMArena human preference rankings guide your selection, as it places #2 (score 14.7) on the agentic arena. [8]

## 5. Claude Fable 5.1
https://modelmatchbook.com/models/anthropic/claude-fable-5.1-20260831

Ranks #3 on LMArena's agentic arena with score 14.2 while placing #5 on Terminal-Bench 4.0 at 57.88%, showing balanced but not leading performance in both evaluations.

**Best when:** Select for multi-step tool-use workflows where LMArena preference matters, ranking #3 (score 14.2) on the agentic arena. [9]

## 6. GLM 5.3
https://modelmatchbook.com/models/z-ai/glm-5.3-20260816

Best open-weight Terminal-Bench result at 41.82% (#9) with Claude Code, though LMArena agentic preference is weak at #25, indicating open-weight tradeoffs for terminal versus open-ended tasks.

**Best when:** Select as the open-weight option for terminal automation, scoring 41.82% on Terminal-Bench 4.0 with Claude Code, the highest open-weight result in that evaluation. [5]

**Watch out for:**
- Expect poor human preference on agentic tasks, ranking #25 (score 1.4) on LMArena's agentic arena. [6]

## 7. GLM 5.3 Flash
https://modelmatchbook.com/models/z-ai/glm-5.3-flash-20260826

GLM 5.3 Flash scores 35.76% ± 3.54 (95% interval) on Terminal-Bench 4.0 (#12 of 23) at none reasoning effort, using Claude Code, measuring a model-and-agent configuration on reproducible terminal tasks.

**Best when:** Its strongest result for this task is on Terminal-Bench, which measures completing difficult repository and terminal tasks through an agent harness. [10]

## 8. Grok 4.7
https://modelmatchbook.com/models/x-ai/grok-4.7-20260916

Mid-tier Terminal-Bench performance at 37.58% (#10) with stronger relative standing on LMArena agentic arena at #16, suggesting Grok Build integration affects terminal scoring.

**Best when:** Evaluate if your deployment uses Grok Build, which achieved 37.58% on Terminal-Bench 4.0 at xhigh reasoning effort. [11]

## FAQ

**What is the best model for Goose?**

Claude Opus 5.5 ranks first as of October 22, 2026. Leads both Terminal-Bench 4.0 and LMArena's agentic arena, scoring 64.85% on reproducible terminal tasks and ranking #1 with a 20.9 score on human preference for tool-use. [1,2]

**What is the best alternative to Claude Opus 5.5?**

Claude Sonnet 5.5 ranks second. Strong second-place showing on Terminal-Bench 4.0 at 61.82%, though its LMArena agentic rank drops to #5, suggesting solid terminal skills with more variable multi-step performance. [3,4]

**Which open-weight model ranks highest for Goose?**

GLM 5.3 is the highest-ranked open-weight model, at #6. Best open-weight Terminal-Bench result at 41.82% (#9) with Claude Code, though LMArena agentic preference is weak at #25, indicating open-weight tradeoffs for terminal versus open-ended tasks. [5,6]

## Sources
1. "Within the 23 catalog models matched to this evaluation: Scores 64.85% ± 3.11 (95% interval) on Terminal-Bench 4.0 (#1 of 23) at max reasoning effort, using Claude Code, measuring a model-and-agent configuration on reproducible terminal tasks." (Terminal-Bench, https://www.tbench.ai/)
2. "Within the 48 catalog models matched to this evaluation: Ranks #1 of 48 on LMArena's agentic arena (score 20.9), measuring tool-use and multi-step task performance from human preference." (LMArena agentic arena, https://lmarena.ai/leaderboard)
3. "Within the 23 catalog models matched to this evaluation: Scores 61.82% ± 2.94 (95% interval) on Terminal-Bench 4.0 (#2 of 23) at max reasoning effort, using Claude Code, measuring a model-and-agent configuration on reproducible terminal tasks." (Terminal-Bench, https://www.tbench.ai/)
4. "Within the 48 catalog models matched to this evaluation: Ranks #5 of 48 on LMArena's agentic arena (score 12.7), measuring tool-use and multi-step task performance from human preference." (LMArena agentic arena, https://lmarena.ai/leaderboard)
5. "Within the 23 catalog models matched to this evaluation: Scores 41.82% ± 3.23 (95% interval) on Terminal-Bench 4.0 (#9 of 23) at max reasoning effort, using Claude Code, measuring a model-and-agent configuration on reproducible terminal tasks." (Terminal-Bench, https://www.tbench.ai/)
6. "Within the 48 catalog models matched to this evaluation: Ranks #25 of 48 on LMArena's agentic arena (score 1.4), measuring tool-use and multi-step task performance from human preference." (LMArena agentic arena, https://lmarena.ai/leaderboard)
7. "Within the 23 catalog models matched to this evaluation: Scores 58.18% ± 3.11 (95% interval) on Terminal-Bench 4.0 (#3 of 23) at max reasoning effort, using Codex, measuring a model-and-agent configuration on reproducible terminal tasks." (Terminal-Bench, https://www.tbench.ai/)
8. "Within the 48 catalog models matched to this evaluation: Ranks #2 of 48 on LMArena's agentic arena (score 14.7), measuring tool-use and multi-step task performance from human preference." (LMArena agentic arena, https://lmarena.ai/leaderboard)
9. "Within the 48 catalog models matched to this evaluation: Ranks #3 of 48 on LMArena's agentic arena (score 14.2), measuring tool-use and multi-step task performance from human preference." (LMArena agentic arena, https://lmarena.ai/leaderboard)
10. "Within the 23 catalog models matched to this evaluation: Scores 35.76% ± 3.54 (95% interval) on Terminal-Bench 4.0 (#12 of 23) at none reasoning effort, using Claude Code, measuring a model-and-agent configuration on reproducible terminal tasks." (Terminal-Bench, https://www.tbench.ai/)
11. "Within the 23 catalog models matched to this evaluation: Scores 37.58% ± 3.54 (95% interval) on Terminal-Bench 4.0 (#10 of 23) at xhigh reasoning effort, using Grok Build, measuring a model-and-agent configuration on reproducible terminal tasks." (Terminal-Bench, https://www.tbench.ai/)