Claude Sonnet 5.5 vs Gemini 4 Argon
| Benchmark | Sonnet 5.5 | Gemini 4 Argon |
|---|---|---|
| Humanity's Last Exam | 45.8 | 57.1 |
| Arena Text · overall | 1,475 · xhigh | 1,524 |
| Arena Text · Chinese | 1,536.9 · xhigh | 1,608.3 |
| Arena Text · English | 1,484.5 · xhigh | 1,537.7 |
| Arena Text · instruction following | 1,488.5 · xhigh | 1,535.6 |
| Arena Text · non-English | 1,454.2 · xhigh | 1,521.7 |
| Arena Text · Russian | 1,445.2 · xhigh | 1,534 |
| Arena Text · Spanish | 1,450.2 · xhigh | 1,504.4 |
| AA Omniscience Index | 21 | 42.4 |
| SimpleQA Verified v1 | 46.5 · max | 76 |
| CritPt (physics) | 24.6 | 27.1 |
| AA Intelligence Index 4.3.2 | 46.8 | 52.6 |
| AA Omniscience · Accuracy | 52 | 49.9 |
| AA Omniscience · Hallucination Rate | 64.6 | 15.1 |
| AA-LCR (long context) 1.1 | 78 | 79.7 |
| GDP.pdf · AA all-pass | 25.2 | 21.8 |
| Chess Puzzles | 59 · max | 59 |
| FrontierMath Tier 4 (v2) | 80.5 · max | 78 |
| FrontierMath Tiers 1–3 (v2) | 88.8 · max | 86 |
| Mystery Game Puzzles | 65 · max | 64 |
| Vals MedCode | 52.9 | 58.8 |
| Vals MedScribe | 91.1 | 87.4 |
| Vals Legal Research Bench | 48.1 | 54.8 |
| Vals Science Index | 59.1 | 55.4 |
| Arena Text · French | 1,496.9 · xhigh | — |
| MLCR-AA | 75 · max | — |
| ARC-AGI-1 · public evaluation | — | 99.5 |
| ARC-AGI-1 · semi-private | — | 98.5 |
| ARC-AGI-2 · public evaluation | — | 97.1 |
| ARC-AGI-2 · semi-private | — | 91.3 |
| ARC-AGI-3 · semi-private | — | 49.8 · low |
| FrontierMath Erdos | 2.9 · max | — |
| MathArena · ArXivMath · August 2026 | — | 82.5 |
| MathArena · ArXivMath · September 2026 | — | 87.7 |
| MathArena · BrokenArXiv · August 2026 | — | 54.2 |
| MathArena · BrokenArXiv · September 2026 | — | 49.7 |
| MathArena · ArXivLean · September 2026 | — | 39.6 |
| Vals LegalBench | — | 88.3 |
Specs
| Spec | Sonnet 5.5 | Gemini 4 Argon |
|---|---|---|
| Input, per 1M | $2 | !$2 |
| Output, per 1M | $10 | !$10 |
| Cost per task? | $0.88 | $1.99 |
| Speed, tok/s | 106 | 84 |
| Context | 1M | 1M |
| Weights | Closed | Closed |
| Released | Sep 2026 | Oct 2026 |
| See images | ✓ | ✓ |
| Read PDFs | ✓ | ✓ |
| Hear audio | — | ✓ |
| Watch video | — | ✓ |
| Call tools | ✓ | ✓ |
| Force a tool call | — | Partial |
| Return strict JSON | ✓ | ✓ |
| Search the web | ✓ | ✓ |
| Run code | ✓ | ✓ |
| Use a computer | ✓ | ? |
What breaks when you switch
Synchronous output stops at 65,536›
With background false, output limits above 65,536 tokens return HTTP 400; stay at or below that limit.
Sampling parameters are ignored›
temperature, top_p and top_k are ignored; remove them from requests.
No explicit caching›
Explicit caching is unavailable; omit explicit cache requests.
No forced tool use›
Forced tool_choice any or tool returns HTTP 400; remove the forced choice.
Replace disabled thinking›
Disabled thinking returns HTTP 400; use between_tools for the lowest thinking setting.
No manual thinking budgets›
Manual thinking budgets and extra between_tools fields return HTTP 400; remove manual budgets and send between_tools without display, budget_tokens or block_binding.
Remove sampling parameters›
Non-default temperature, top_p or top_k returns HTTP 400; remove these parameters.
No assistant prefill›
A prefilled final assistant turn returns HTTP 400; remove that prefill.
Old computer tool rejected · Claude API / Google Cloud›
The computer_20251124 tool returns HTTP 400 on the Claude API and Google Cloud; use computer_toolset_20260801 there.