Part of Best LLM for AI Agents

Best LLM for OpenHands

Claude Fable 5.2 ranks first for OpenHands as of November 15, 2026.[1][2][3][4] Leads both Terminal-Bench 4.1 and LMArena's agentic arena, scoring 65.76% and ranking #1 respectively, indicating strong terminal task execution and human-preferred tool use. Claude Opus 5.5 is the next-ranked alternative. Matches the top Terminal-Bench 4.1 score of 65.76% but ranks #2 on LMArena's agentic arena, showing equivalent terminal capability with slightly lower human preference for tool use.

About this recommendation

Updated
Evidence through
Sources
15
Revision
v125

Decision audit

Why this result

Inspect the inputs and the computed order behind the recommendation.

Models screened

24

live candidates

Evaluation feeds

5

task-weighted

Winner coverage

56%

intended feed weight

Largest provider share

3 of 8

Anthropic

Established source breadth. 2 citation families and 0 practitioner families provide attributed capability reports; 0 cautionary threads retained. The largest citation family contributes 53%.

Anthropic: Claude Fable 5.2 leads the measured task comparison with 2 independent evaluation families. Community volume, release age and usage popularity add no ranking points. This category also weights its stated price or context criteria.

Real-world track record is still developing. The recommendation can lead on independent measurements while practitioner reports accumulate.

Sources evaluated

These are the weights used in scoring. Missing measurements are omitted from each model’s average. A feed with no result for any model released in the last 60 days is stale and does not score until it catches up.

winner: Claude Fable 5.2
Evaluation feedWeightWinner resultField measured
Berkeley Function Callingstale
0%
no current result0/24
LMArena AgentObserved 2026-11-11 · evaluation date not supplied · max effort
43%
#124/24
price weight
28%
16/10024/24
Terminal-Bench (4.1)Observed 2026-11-15 · evaluated 2026-11-13 · Claude Code; max effort · 95% interval ±2.85
28%
#124/24
SWE-rebenchstale
0%
no current result0/24

Provider concentration

Each exact model is scored separately; provider identity is not a ranking input.

Anthropic38%
  • Anthropic3 models
  • OpenAI2 models
  • deepseek1 model
  • Google1 model
  • Moonshot AI1 model

Decision table

Every published model is shown in computed order. Practitioner sources are distinct community threads, not the citations repeated in the prose below.

  1. Anthropic · coverage 56%

    #1 LMArena Agent · #1 Terminal-Bench

  2. Anthropic · coverage 56%

    #1 Terminal-Bench · #2 LMArena Agent

  3. Google · coverage 56%

    #3 LMArena Agent · #3 Terminal-Bench

  4. Anthropic · coverage 56%

    #4 LMArena Agent · #4 Terminal-Bench

  5. Moonshot AI · coverage 56%

    #7 Terminal-Bench · #14 LMArena Agent

  6. OpenAI · coverage 56%

    #5 Terminal-Bench · #7 LMArena Agent

  7. OpenAI · coverage 56%

    #8 LMArena Agent · #11 Terminal-Bench

  8. deepseek · coverage 56%

    #19 Terminal-Bench · #21 LMArena Agent

Task score averages the available, task-weighted measurements. Benchmark positions compare models matched to our catalog. Models measured by at least two independent evaluator families are considered first. Missing tests, practitioner report counts, popularity and release age add no points. Coverage describes evidence breadth, not a probability of correctness. Scores within two points are shown as close under this policy; this is not a statistical confidence interval. When the leading scores are that close, the model measured by more evaluator families ranks first.

  1. 01

    Claude Fable 5.2Anthropic · $8 in / $40 out per M tokens · 1M context

    Leads both Terminal-Bench 4.1 and LMArena's agentic arena, scoring 65.76% and ranking #1 respectively, indicating strong terminal task execution and human-preferred tool use.

    Best when: Deploy when human preference for multi-step tool use is the priority, as it ranks #1 on LMArena's agentic arena with score 30.4.
    Source 2
    “Within the 52 catalog models matched to this evaluation: Ranks #1 of 52 on LMArena's agentic arena (score 30.4), measuring tool-use and multi-step task performance from human preference.”
    LMArena agentic arenaOpen original ↗

    No practitioner reports are linked yet, so no cautions are listed. Absence of reports is not evidence of reliability.

  2. 02

    Claude Opus 5.5Anthropic · $4 in / $20 out per M tokens · 1M context

    Matches the top Terminal-Bench 4.1 score of 65.76% but ranks #2 on LMArena's agentic arena, showing equivalent terminal capability with slightly lower human preference for tool use.

    Best when: Its strongest result for this task is on Terminal-Bench, which measures completing difficult repository and terminal tasks through an agent harness.
    Source 3
    “Within the 46 catalog models matched to this evaluation: Scores 65.76% ± 3.11 (95% interval) on Terminal-Bench 4.1 (#1 of 46) at max reasoning effort, using Claude Code, measuring a model-and-agent configuration on reproducible terminal tasks.”
    Terminal-BenchOpen original ↗

    Watch out for

    • Expect lower human preference scores on multi-step tool use compared to the top-ranked model, as it ranks #2 on LMArena's agentic arena with score 19.5 versus 30.4.
      Source 4
      “Within the 52 catalog models matched to this evaluation: Ranks #2 of 52 on LMArena's agentic arena (score 19.5), measuring tool-use and multi-step task performance from human preference.”
      LMArena agentic arenaOpen original ↗
  3. 03

    Gemini 4 CarbonGoogle · $2 in / $10 out per M tokens · 1.0M context

    Ranks #3 on both Terminal-Bench 4.1 (64.55%) and LMArena's agentic arena, demonstrating strong but slightly lower performance than the top two on terminal tasks and tool use.

    Best when: Its strongest result for this task is on Terminal-Bench, which measures completing difficult repository and terminal tasks through an agent harness.
    Source 6
    “Within the 46 catalog models matched to this evaluation: Scores 64.55% ± 3.23 (95% interval) on Terminal-Bench 4.1 (#3 of 46) at high reasoning effort, using Antigravity CLI, measuring a model-and-agent configuration on reproducible terminal tasks.”
    Terminal-BenchOpen original ↗

    Watch out for

    • Expect lower human preference on multi-step tool use than top models, ranking #3 on LMArena's agentic arena with score 17.
      Source 7
      “Within the 52 catalog models matched to this evaluation: Ranks #3 of 52 on LMArena's agentic arena (score 17), measuring tool-use and multi-step task performance from human preference.”
      LMArena agentic arenaOpen original ↗
  4. 04

    Claude Sonnet 5.5Anthropic · $2 in / $10 out per M tokens · 1M context

    Ranks #4 on both Terminal-Bench 4.1 (63.94%) and LMArena's agentic arena, showing consistent fourth-place performance on terminal tasks and tool use.

    Best when: Its strongest result for this task is on Terminal-Bench, which measures completing difficult repository and terminal tasks through an agent harness.
    Source 8
    “Within the 46 catalog models matched to this evaluation: Scores 63.94% ± 3 (95% interval) on Terminal-Bench 4.1 (#4 of 46) at max reasoning effort, using Claude Code, measuring a model-and-agent configuration on reproducible terminal tasks.”
    Terminal-BenchOpen original ↗

    Watch out for

    • Expect lower human preference on multi-step tool use than top models, ranking #4 on LMArena's agentic arena with score 16.
      Source 9
      “Within the 52 catalog models matched to this evaluation: Ranks #4 of 52 on LMArena's agentic arena (score 16), measuring tool-use and multi-step task performance from human preference.”
      LMArena agentic arenaOpen original ↗
  5. 05

    Kimi K3.1 ProMoonshot AI · $0.24 in / $8.40 out per M tokens · 1.0M context · open-weight

    Kimi K3.1 Pro scores 59.7% ± 4.18 (95% interval) on Terminal-Bench 4.1 (#7 of 46) at max 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.
    Source 5
    “Within the 46 catalog models matched to this evaluation: Scores 59.7% ± 4.18 (95% interval) on Terminal-Bench 4.1 (#7 of 46) at max reasoning effort, using Claude Code, measuring a model-and-agent configuration on reproducible terminal tasks.”
    Terminal-BenchOpen original ↗

    No practitioner reports are linked yet, so no cautions are listed. Absence of reports is not evidence of reliability.

  6. 06

    GPT-6.1 SolOpenAI · $2 in / $10 out per M tokens · 1.1M context

    Ranks #5 on Terminal-Bench 4.1 (60.91%) and #7 on LMArena's agentic arena, showing solid terminal performance with moderate tool use preference.

    Best when: Its strongest result for this task is on Terminal-Bench, which measures completing difficult repository and terminal tasks through an agent harness.
    Source 10
    “Within the 46 catalog models matched to this evaluation: Scores 60.91% ± 2.97 (95% interval) on Terminal-Bench 4.1 (#5 of 46) at max reasoning effort, using Codex, measuring a model-and-agent configuration on reproducible terminal tasks.”
    Terminal-BenchOpen original ↗

    Watch out for

    • Expect moderate human preference on multi-step tool use, ranking #7 on LMArena's agentic arena with score 12.5.
      Source 11
      “Within the 52 catalog models matched to this evaluation: Ranks #7 of 52 on LMArena's agentic arena (score 12.5), measuring tool-use and multi-step task performance from human preference.”
      LMArena agentic arenaOpen original ↗
  7. 07

    GPT-6.1 TerraOpenAI · $1 in / $5 out per M tokens · 1.1M context

    Ranks #11 on Terminal-Bench 4.1 (53.33%) and #8 on LMArena's agentic arena, showing below-average terminal performance but slightly better tool use preference.

    Best when: Use when human preference for multi-step tool use matters more than raw terminal accuracy, ranking #8 on LMArena's agentic arena with score 12.
    Source 12
    “Within the 52 catalog models matched to this evaluation: Ranks #8 of 52 on LMArena's agentic arena (score 12), measuring tool-use and multi-step task performance from human preference.”
    LMArena agentic arenaOpen original ↗

    Watch out for

    • Expect below-average terminal task performance, scoring 53.33% on Terminal-Bench 4.1 (#11 of 46) using Codex.
      Source 13
      “Within the 46 catalog models matched to this evaluation: Scores 53.33% ± 3.09 (95% interval) on Terminal-Bench 4.1 (#11 of 46) at max reasoning effort, using Codex, measuring a model-and-agent configuration on reproducible terminal tasks.”
      Terminal-BenchOpen original ↗
  8. 08

    DeepSeek V4.1 Flash 1017deepseek · $0.020 in / $0.11 out per M tokens · 1.0M context · open-weight

    An open-weight option ranking #19 on Terminal-Bench 4.1 (38.48%) and #21 on LMArena's agentic arena, showing limited capability on both metrics.

    Best when: Use as an open-weight fallback for terminal tasks, scoring 38.48% on Terminal-Bench 4.1 (#19 of 46).
    Source 14
    “Within the 46 catalog models matched to this evaluation: Scores 38.48% ± 3.76 (95% interval) on Terminal-Bench 4.1 (#19 of 46) at max reasoning effort, using Claude Code, measuring a model-and-agent configuration on reproducible terminal tasks.”
    Terminal-BenchOpen original ↗

    Watch out for

    • Expect limited human preference on multi-step tool use, ranking #21 on LMArena's agentic arena with score 2.
      Source 15
      “Within the 52 catalog models matched to this evaluation: Ranks #21 of 52 on LMArena's agentic arena (score 2), measuring tool-use and multi-step task performance from human preference.”
      LMArena agentic arenaOpen original ↗

Frequently asked

What is the best model for OpenHands?
Claude Fable 5.2 ranks first as of November 15, 2026. Leads both Terminal-Bench 4.1 and LMArena's agentic arena, scoring 65.76% and ranking #1 respectively, indicating strong terminal task execution and human-preferred tool use.[1][2]
What is the best alternative to Claude Fable 5.2?
Claude Opus 5.5 ranks second. Matches the top Terminal-Bench 4.1 score of 65.76% but ranks #2 on LMArena's agentic arena, showing equivalent terminal capability with slightly lower human preference for tool use.[3][4]
Which open-weight model ranks highest for OpenHands?
Kimi K3.1 Pro is the highest-ranked open-weight model, at #5. Kimi K3.1 Pro scores 59.7% ± 4.18 (95% interval) on Terminal-Bench 4.1 (#7 of 46) at max reasoning effort, using Claude Code, measuring a model-and-agent configuration on reproducible terminal tasks.[5]

Sources

  1. 1

    “Within the 46 catalog models matched to this evaluation: Scores 65.76% ± 2.85 (95% interval) on Terminal-Bench 4.1 (#1 of 46) at max reasoning effort, using Claude Code, measuring a model-and-agent configuration on reproducible terminal tasks.”

    Terminal-Bench · Benchmark · Nov 15, 2026
  2. 2

    “Within the 52 catalog models matched to this evaluation: Ranks #1 of 52 on LMArena's agentic arena (score 30.4), measuring tool-use and multi-step task performance from human preference.”

    LMArena agentic arena · Benchmark · Nov 11, 2026
  3. 3

    “Within the 46 catalog models matched to this evaluation: Scores 65.76% ± 3.11 (95% interval) on Terminal-Bench 4.1 (#1 of 46) at max reasoning effort, using Claude Code, measuring a model-and-agent configuration on reproducible terminal tasks.”

    Terminal-Bench · Benchmark · Nov 15, 2026
  4. 4

    “Within the 52 catalog models matched to this evaluation: Ranks #2 of 52 on LMArena's agentic arena (score 19.5), measuring tool-use and multi-step task performance from human preference.”

    LMArena agentic arena · Benchmark · Nov 11, 2026
  5. 5

    “Within the 46 catalog models matched to this evaluation: Scores 59.7% ± 4.18 (95% interval) on Terminal-Bench 4.1 (#7 of 46) at max reasoning effort, using Claude Code, measuring a model-and-agent configuration on reproducible terminal tasks.”

    Terminal-Bench · Benchmark · Nov 15, 2026
  6. 6

    “Within the 46 catalog models matched to this evaluation: Scores 64.55% ± 3.23 (95% interval) on Terminal-Bench 4.1 (#3 of 46) at high reasoning effort, using Antigravity CLI, measuring a model-and-agent configuration on reproducible terminal tasks.”

    Terminal-Bench · Benchmark · Nov 15, 2026
  7. 7

    “Within the 52 catalog models matched to this evaluation: Ranks #3 of 52 on LMArena's agentic arena (score 17), measuring tool-use and multi-step task performance from human preference.”

    LMArena agentic arena · Benchmark · Nov 11, 2026
  8. 8

    “Within the 46 catalog models matched to this evaluation: Scores 63.94% ± 3 (95% interval) on Terminal-Bench 4.1 (#4 of 46) at max reasoning effort, using Claude Code, measuring a model-and-agent configuration on reproducible terminal tasks.”

    Terminal-Bench · Benchmark · Nov 15, 2026
  9. 9

    “Within the 52 catalog models matched to this evaluation: Ranks #4 of 52 on LMArena's agentic arena (score 16), measuring tool-use and multi-step task performance from human preference.”

    LMArena agentic arena · Benchmark · Nov 11, 2026
  10. 10

    “Within the 46 catalog models matched to this evaluation: Scores 60.91% ± 2.97 (95% interval) on Terminal-Bench 4.1 (#5 of 46) at max reasoning effort, using Codex, measuring a model-and-agent configuration on reproducible terminal tasks.”

    Terminal-Bench · Benchmark · Nov 15, 2026
  11. 11

    “Within the 52 catalog models matched to this evaluation: Ranks #7 of 52 on LMArena's agentic arena (score 12.5), measuring tool-use and multi-step task performance from human preference.”

    LMArena agentic arena · Benchmark · Nov 11, 2026
  12. 12

    “Within the 52 catalog models matched to this evaluation: Ranks #8 of 52 on LMArena's agentic arena (score 12), measuring tool-use and multi-step task performance from human preference.”

    LMArena agentic arena · Benchmark · Nov 11, 2026
  13. 13

    “Within the 46 catalog models matched to this evaluation: Scores 53.33% ± 3.09 (95% interval) on Terminal-Bench 4.1 (#11 of 46) at max reasoning effort, using Codex, measuring a model-and-agent configuration on reproducible terminal tasks.”

    Terminal-Bench · Benchmark · Nov 15, 2026
  14. 14

    “Within the 46 catalog models matched to this evaluation: Scores 38.48% ± 3.76 (95% interval) on Terminal-Bench 4.1 (#19 of 46) at max reasoning effort, using Claude Code, measuring a model-and-agent configuration on reproducible terminal tasks.”

    Terminal-Bench · Benchmark · Nov 15, 2026
  15. 15

    “Within the 52 catalog models matched to this evaluation: Ranks #21 of 52 on LMArena's agentic arena (score 2), measuring tool-use and multi-step task performance from human preference.”

    LMArena agentic arena · Benchmark · Nov 11, 2026

Rankings synthesized from community evidence and open benchmarks. See methodology. Not driven by vendor marketing.