Recommendation for Autonomous workflows
Autonomous Agents
Our top recommendation for Autonomous Agents, based on the public evidence we track, is Anthropic: Claude Fable 5.[1][2][3][4] Assign to plan-review, code-review, and security advisory roles where taxonomy guidelines explicitly recommend Fable 5's 'ultra' tier for verification-heavy work. Watch out: Accept lower raw benchmark scores: it resolves only 34.05% on Frontier-Bench versus GPT-5.6 Sol's 53.6 on Agents' Last Exam, and costs roughly 4x more at comparable reasoning levels. Z.ai: GLM 5.2 is the next-ranked alternative. Deploy on self-hosted infrastructure with 36GB GPUs for 250K context windows when API rate limits or subscription costs block commercial alternatives.
About this recommendation
- Updated
- Sep 25, 2026
- Evidence through
- Sep 25, 2026
- Sources
- 14
- Revision
- v80
Decision audit
Why this result
Inspect the inputs and the computed order behind the recommendation.
Models screened
20
live candidates
Evaluation feeds
7
task-weighted
Winner coverage
99%
intended feed weight
Largest provider share
2 of 3
Anthropic
Sources evaluated
The task sets these weights before any model is scored.
| Evaluation feed | Weight | Winner result | Field measured |
|---|---|---|---|
| Terminal-Bench 2.1 | 28% | #12 | 14/20 |
| SWE-rebench | 24% | #1 | 14/20 |
| Frontier-Bench | 18% | #7 | 11/20 |
| LiveBench Agentic Coding | 12% | #7 | 17/20 |
| LMArena Agent | 8% | #2 | 14/20 |
| OpenRouter usage | 5% | 79/100 | 20/20 |
| OSWorld Verified | 5% | not measured | 4/20 |
Provider concentration
Each exact model is scored separately; provider identity is not a ranking input.
- Anthropic2 models
- Z.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.
| Rank | Model | Relative score | Coverage | Practitioner evidence | Strongest measured reason |
|---|---|---|---|---|---|
| 01 | Claude Fable 5Anthropic | 78 | 99% | 7 threads · 5 families · 0 cautions | #1 SWE-rebench · #2 LMArena Agent |
| 02 | GLM 5.2Z.ai | 64 | 86% | 8 threads · 6 families · 5 cautions | #9 SWE-rebench · #11 LMArena Agent |
| 03 | Claude Opus 5.5Anthropic | 49 | 25% | 3 threads · 3 families · 1 cautions | #2 LiveBench Agentic Coding |
Relative score combines normalized benchmark quality and signal coverage; independent practitioner evidence and freshness are bounded tie-breakers. It is an ordering score, not an absolute quality percentage. The writing model receives this order and cannot change it.
Claude Fable 5 ranks #2 on LMArena's agentic arena and excels at plan-review and advisory roles in multi-tier agent architectures, though it trails GPT-5.6 Sol on Agents' Last Exam by a significant margin.
Best when: Assign to plan-review, code-review, and security advisory roles where taxonomy guidelines explicitly recommend Fable 5's 'ultra' tier for verification-heavy work.
Tips
- Assign to plan-review, code-review, and security advisory roles where taxonomy guidelines explicitly recommend Fable 5's 'ultra' tier for verification-heavy work.
- Use for long-running game and web-app builds where it demonstrates solid autonomous execution from natural language specifications.
- Deploy as a monitoring layer over cheaper executors: it can validate execution, create adjustment tasks, and maintain quality without hitting rate limits.
Watch out for
- Accept lower raw benchmark scores: it resolves only 34.05% on Frontier-Bench versus GPT-5.6 Sol's 53.6 on Agents' Last Exam, and costs roughly 4x more at comparable reasoning levels.
- Watch for workflow model pinning: subagents spawn with the session's starting model, not the current model, so mid-session `/model` switches never reach workflow agents.
GLM 5.2 offers an open-weight alternative with broad deployment route availability and solid performance on near-Sonnet level tasks, though it ranks in the bottom quartile on Frontier-Bench and exhibits token consumption issues.
Best when: Deploy on self-hosted infrastructure with 36GB GPUs for 250K context windows when API rate limits or subscription costs block commercial alternatives.
Tips
- Deploy on self-hosted infrastructure with 36GB GPUs for 250K context windows when API rate limits or subscription costs block commercial alternatives.
- Use as a fallback in model routing chains where it serves as a viable backup before escalating to more expensive closed-weight options.
Watch out for
- Accept minimal Frontier-Bench performance: it resolves only 4.59% of tasks, ranking last among 16 public submissions.
- Watch for sporadic Chinese character output and higher-than-expected token consumption that can burn through quotas quickly.
- Expect lower human preference rankings: it places #10 of 36 on LMArena's agentic arena with a score of 4.4.
Claude Opus 5.5 scores #2 on LiveBench Agentic Coding and demonstrates strong self-correction capabilities in multi-agent orchestration, though it requires turn limits to prevent runaway loops.
Best when: Use for complex multi-file changes where it can judge subagent limitations, terminate misperforming workers, and reschedule with appropriate capability tiers.
Tips
- Use for complex multi-file changes where it can judge subagent limitations, terminate misperforming workers, and reschedule with appropriate capability tiers.
- Deploy for extended sessions with 1M context windows where it maintains coherent state across thousands of turns with merge subagents and auditors.
Watch out for
- Implement hard turn caps: without limits, the model can loop indefinitely, requiring manual intervention after 38+ turns.
Frequently asked
- What is the top-ranked model for Autonomous Agents?
- Anthropic: Claude Fable 5 ranks first in the current evidence-weighted comparison. Assign to plan-review, code-review, and security advisory roles where taxonomy guidelines explicitly recommend Fable 5's 'ultra' tier for verification-heavy work.[1]
- What should I watch out for with Anthropic: Claude Fable 5?
- Accept lower raw benchmark scores: it resolves only 34.05% on Frontier-Bench versus GPT-5.6 Sol's 53.6 on Agents' Last Exam, and costs roughly 4x more at comparable reasoning levels.[2][3]
- What is an alternative to Anthropic: Claude Fable 5?
- Z.ai: GLM 5.2 is the next-ranked option. Deploy on self-hosted infrastructure with 36GB GPUs for 250K context windows when API rate limits or subscription costs block commercial alternatives.[4]
Sources
- 1
“## Summary A five-model advisor review of the newly merged model-tier taxonomy (`c266410`, agent-harness#309) converged on an author/reviewer split, but rejected making the heavy tier the default implementation tier. Recommended upstream policy: - roadmap/plan/supervise: Claude Opus 5 (`heavy`) - execute/repair: Claude Sonnet 5 (`regular`), with a recorded escalation to Opus 5 after repeated verification failure - plan-review/code-review/advise/security: Claude Fable 5 (`ultra`) - worker/econom…”
ViperJuice · GitHub · Jul 26, 2026 - 2
“using Claude Code at max effort, resolves 34.05% ± 1.71 of Frontier-Bench tasks (#7 of 16 public submissions), measuring a submitted model-and-agent configuration on diverse, difficult agent work.”
Frontier-Bench · Benchmark · Jul 13, 2026 - 3
“"On Agents’ Last Exam (opens in a new window), an evaluation of long-running professional workflows across 55 fields, GPT‑5.6 Sol sets a new high of 53.6, eclipsing Claude Fable 5 (adaptive reasoning) by 13.1 points. Even at medium reasoning, it beats Fable 5 by 11.4 points at roughly one-quarter the estimated cost. That efficiency extends to smaller models, which are essential to making intelligence more abundant and affordable: GPT‑5.6 Terra and GPT‑5.6 Luna outperform Fable 5 at around one-s…”
saberience · Hacker News · Jul 9, 2026 - 4
“Next to my Claude Pro plan, I have subbed to OpenCode Go. I find the OpenCode UX much better than in Claude Code CLI. As for models, I started a few months ago with GLM 5.1 and it was solid and could archive near sonnet-level tasks. It weirdly sputtered out Chinese characters sometimes. Then I switched to Kimi K2.6, which is the Chinese model I used the most until now. It used way too many reasoning tokens (improved in k2.7). But executed Claude created plans reliably. Now I’m back with GLM 5.2…”
michimagdesign · Hacker News · Jun 24, 2026 - 5
“<p>Back in 2022 <a href="https://twitter.com/simonw/status/1555626060384911360">I tweeted</a> screenshots of a game concept generated by GPT-3 and some concept "art" created using DALL-E. Today, on the fourth anniversary of that tweet, I decided to see if Claude Fable 5 (running in <a href="https://code.claude.com/docs/en/claude-code-on-the-web">Claude Code for web</a>) could build the entire game from the content of that tweet. It did a pretty good job of it!</p> <p>You can <a href="https://si…”
Engineering publication · Aug 5, 2026 - 6
“I have started testing roles separation: -- Stronger model from claude subscription is responsible for analysis solution design task creation. -- Cheaper model (qwen3.6, local deployment on 36GB GPU with 250K context size) is responsible for task execution Doing this for less then 24 hours. Fable can monitor and validate execution and create smaller tasks when something has to be adjusted after qwen execution. It is slower ofc, but this can work non stop without hitting the limits even when usi…”
sermakarevich · Hacker News · Jun 12, 2026 - 7
“## Problem Workflow subagents run on the model the session **started with**, not the model the session is **currently on**. Switching models mid-session with `/model` never reaches workflow agent spawning. Evidence from a live fleet session: 1. Started a pi session with model `aigw/round-robin/qwen3.8-max`. 2. Switched to `dongli/claude-fable-5` via `/model` (~1h later). 3. Ran the built-in `code-review` workflow. All 7 agents spawned with `aigw/round-robin/qwen3.8-max` (recorded in the run fil…”
Explorer1092 · GitHub · Aug 11, 2026 - 8
“An operator switched the model to GLM-5.2 in the web console, asked the agent what it was running, and was told: > I'm running **GPT-5.6 Sol** (`openai-codex:gpt-5.6-sol`) as the primary model, configured at `effort: high`. > If that model is unavailable I fall back through a chain: Gemini 3.1 Pro → Gemini 3.5 Flash → Kimi K2.7 Code → GLM-5.2 → Claude Opus 4.8 → Claude Fable 5 → and a few others, ending with GPT-5.6 Terra. The override had worked. The run record for that exact turn: GLM-5.2 ser…”
robertsreberski · GitHub · Jul 25, 2026 - 9
“using Claude Code at max effort, resolves 4.59% ± 0.97 of Frontier-Bench tasks (#16 of 16 public submissions), measuring a submitted model-and-agent configuration on diverse, difficult agent work.”
Frontier-Bench · Benchmark · Jul 19, 2026 - 10
“Since I quit my Claude subscription, every month I spend $20 (the cost of CC pro plan) playing around with new models and new providers. Currently testing M3 for agentic tasks. It works OK and their token plan is very cheap. Highly recommend for claw hermes type of work. Tested GLM 5.1 for coding last month and it burned through my tokens a bit too quickly, but it worked well enough.”
throwa356262 · Hacker News · Jun 20, 2026 - 11
“Ranks #10 of 36 on LMArena's agentic arena (score 4.4), measuring tool-use and multi-step task performance from human preference.”
LMArena agentic arena · Benchmark · Sep 15, 2026 - 12
“I have to say that I'm very pleased with this version. My prime anecdotal reason: I asked Opus 5.5 to make a sweeping change with a few Sonnet agents. It clarified the scope, we agreed and then it started. I saw some problems on the way and asked it. It's answer: This didn't go as planned. The Sonnet agents don't have the right judgement capacity for this, so they take shortcuts based on their limited scope X. I suggest we terminate them, roll back their changes and I can schedule an Opus agent…”
hgo · Hacker News · Sep 23, 2026 - 13
“# KAIF 2.7 update field report — QA_Engineer > Deployment: sphere `qa-engineering` · language `ru` (`i18n: translated`) · agents `claude-code` + > `codex` + 3 mirrors · macOS 26.0.1 arm64 · Node v23.11.0 · executed by Claude Opus 5.5 (1M context) in > Claude Code, with four merge subagents, one style subagent, one obligation auditor and one judge, each > in a clean context. Route: **thin `KAIF.md` 2.7 (fresh core) bound to a sandbox rehearsal, 2.4 → 2.7**, > `tracking: origin` throughout. Writt…”
MikalaiKryvusha · GitHub · Sep 23, 2026 - 14
“Following #2: there's no limit on how many turns a run can take, so a model stuck in a loop keeps going until someone restarts the bot. Today's loop was only stopped by hand after 38 turns. A cap, even a generous one, would have ended it on its own. Once the cap is hit, the bot should probably tell the channel it gave up rather than go quiet. 🤖 Claude Opus 5.5”
dcabo · GitHub · Sep 23, 2026
Rankings synthesized from community evidence and open benchmarks. See methodology. Not driven by vendor marketing.