Claude Haiku 5.5 vs GPT-6 Astra
GPT-6 Astra leads Claude Haiku 5.5 by 13.4 points on the Shogo Agent Index, but their ranges overlap, so treat it as close. Claude Haiku 5.5 wins 1 and GPT-6 Astra wins 20 of 21 shared benchmarks. Claude Haiku 5.5 is cheaper for agent workloads.
Add more models to this comparison →Pillars
| Pillar | Claude Haiku 5.5 | GPT-6 Astra |
|---|---|---|
| Agentic coding | 67 | 80 |
| Long-horizon autonomy | 62 | 80 |
| Knowledge-work agents | 46 | 56 |
| Computer use | 47 | 67 |
| Security & context learning | 55 | 63 |
Shared benchmarks
| Benchmark | Claude Haiku 5.5 | GPT-6 Astra |
|---|---|---|
| APEX-Agentsindex | 56.1% | 64.7% |
| Vending-Bench 2index | $3,602 | $15,515 |
| FrontierSWEindex | 43.8% | 65.5% |
| ARC-AGI | 89.0% | 98.5% |
| ARC-AGI-2 | 61.3% | 95.0% |
| Chess Puzzles | 55.0% | 76.0% |
| CritPt | 22.6% | 31.7% |
| DTBench | 95.0% | 97.3% |
| FrontierCode | 46.4% | 53.3% |
| FrontierMath Tier 4 | 46.3% | 97.6% |
| FrontierMath Tiers 1-3 | 75.1% | 93.7% |
| GPQA Diamond | 89.6% | 95.8% |
| Humanity's Last Exam | 30.1% | 54.8% |
| LMCA | 51.1% | 64.4% |
| MirrorCode | 78.9% | 60.0% |
| OTIS Mock AIME | 98.9% | 100.0% |
| ProofBench | 89.0% | 99.0% |
| SciCode | 55.0% | 56.5% |
| SimpleBench | 59.2% | 86.5% |
| SimpleQA Verified | 23.8% | 75.6% |
| WeirdML v3 | 11.7% | 42.2% |
Specs and price
| Claude Haiku 5.5 | GPT-6 Astra | |
|---|---|---|
| Agent Index | 57.5 (#18) | 70.9 (#4) |
| Agent $/M | $0.050 | $4.62 |
| Input / output $/M | $0.10 / $0.50 | $10.00 / $50.00 |
| Context | 1M | 1.1M |
| Released | Oct 7, 2026 | Sep 4, 2026 |
| Weights | Closed | Closed |
| API hosts | 5 | 6 |
Don’t pick one. Route.
Shogo can run Claude Haiku 5.5 and GPT-6 Astra side by side and send each step of an agent to the one that’s cheaper for it.
Try both in Shogo