Model comparison
Claude Sonnet 4.6 vs o3
Head-to-head evidence from 15 shared benchmark results across 7 categories. Overall scores shown here use the public BenchAlign v5 ranking lane.
Public leaderboard positions: Claude Sonnet 4.6 #32 (Supported); o3 #131 (Supported). Intervals and evidence labels describe ranking uncertainty, not a guarantee for a specific workload.
Evidence parity. Claude Sonnet 4.6 and o3 share 15 comparable benchmark results. 1 of 8 categories are comparable. 18 results are unique to Claude Sonnet 4.6; 1 to o3.
Updated July 20, 2026- Shared results
- 15
- Claude Sonnet 4.6 only
- 18
- o3 only
- 1
- Comparable categories
- 1 / 8
Pick Claude Sonnet 4.6 if you want the stronger benchmark profile. o3 only becomes the better choice if you want the cheaper token bill or you want the stronger reasoning-first profile.
Confidence note. This is a partial-evidence comparison with 15 shared benchmark results across 7 evidence categories; 1 of 8 categories currently have scoreable aggregates for both models. Treat the verdict as directional until coverage is more balanced.
Why this result
Claude Sonnet 4.6 is clearly ahead on the BenchAlign aggregate, 65.07 to 47.89. The gap is large enough that you do not need to squint at the spreadsheet to see the difference.
Claude Sonnet 4.6's sharpest advantage is in mathematics, where it averages 26.4 against 14.5. The single biggest benchmark swing on the page is FrontierMath v2 (Tiers 1-3), 32.400% to 18.685%.
Claude Sonnet 4.6 is also the more expensive model on tokens at $3.00 input / $15.00 output per 1M tokens, versus $2.00 input / $8.00 output per 1M tokens for o3. o3 is the reasoning model in the pair, while Claude Sonnet 4.6 is not. That usually helps on harder chain-of-thought-heavy tests, but it can also mean more latency and more token spend in real use.
Category breakdown
Exact category averages are shown below. Not measured means BenchLM does not have enough sourced public coverage for that model and category.
| Category | Claude Sonnet 4.6 | Δ | o3 |
|---|---|---|---|
| Math | Claude Sonnet 4.626.4 | Margin← 11.9 | o314.5 |
| Agentic | Claude Sonnet 4.665.2 | MarginNo overlap | o3Not measured |
| Coding | Claude Sonnet 4.669.1 | MarginNo overlap | o3Not measured |
| Knowledge | Claude Sonnet 4.666.0 | MarginNo overlap | o3Not measured |
| Multimodal | Claude Sonnet 4.677.4 | MarginNo overlap | o3Not measured |
Decisive benchmark drivers
The largest measured benchmark gaps in this matchup, with exact reported values.
More
- Source ↗
FrontierMath v2 (Tiers 1-3)
MathA 32.400%B 18.685%Winner: Claude Sonnet 4.6Δ 13.7FrontierMath v2 (Tiers 1-3): Claude Sonnet 4.6 scored 32.400%; o3 scored 18.685%. Claude Sonnet 4.6 wins this benchmark. - Source ↗
FrontierMath v2 (Tier 4)
MathA 8.300%B 2.083%Winner: Claude Sonnet 4.6Δ 6.2FrontierMath v2 (Tier 4): Claude Sonnet 4.6 scored 8.300%; o3 scored 2.083%. Claude Sonnet 4.6 wins this benchmark.
Operational comparison
Runtime and commercial metrics are compared only when both models have a complete sourced value.
| Metric | Claude Sonnet 4.6 | o3 | Comparison |
|---|---|---|---|
| Input / output priceUSD per 1M tokens | Claude Sonnet 4.6$3 input / $15 output | o3$2 input / $8 output | o3 has the lower combined listed price. |
| Generation speedtokens per second | Claude Sonnet 4.644 tok/s | o3118 tok/s | o3 has the higher measured throughput. |
| First-answer latencyseconds to first token | Claude Sonnet 4.61.48 s | o35.38 s | Claude Sonnet 4.6 reaches the first token sooner. |
| Context windowmaximum listed tokens | Claude Sonnet 4.6200K | o3200K | Listed context windows are equal. |
Benchmark Deep Dive
Agentic8 benchmarks
| Benchmark | Claude Sonnet 4.6 | o3 | Result |
|---|---|---|---|
| Terminal-Bench 2.0Source | 59.1% | — | Not comparable |
| OSWorld-VerifiedSource | 72.1% | — | Not comparable |
| Claw-EvalSource | 67.8% | — | Not comparable |
| CyberGymSource | 65.2% | — | Not comparable |
| τ²-bench resultsSource | 79.5% | 80.7% | o3 leads |
| Gert LabsSource | 62.92% | — | Not comparable |
| OSWorld 2.0Source | 8.3% | — | Not comparable |
| JobBenchSource | 36.9% | — | Not comparable |
Coding7 benchmarks
| Benchmark | Claude Sonnet 4.6 | o3 | Result |
|---|---|---|---|
| SWE-bench VerifiedSource | 79.6% | — | Not comparable |
| SWE-RebenchSource | 60.7% | — | Not comparable |
| React Native EvalsSource | 80.6% | — | Not comparable |
| Vibe Code BenchSource | 51.48% | — | Not comparable |
| cursorBench31Source | 48.8% | — | Not comparable |
| AA-SciCodeSource | 46.9% | 41.0% | Claude Sonnet 4.6 leads |
| FrontierCode 1.1 MainSource | 24.3% | — | Not comparable |
Reasoning2 benchmarks
Knowledge10 benchmarks
| Benchmark | Claude Sonnet 4.6 | o3 | Result |
|---|---|---|---|
| GPQASource | 89.9% | — | Not comparable |
| SuperGPQASource | 95% | — | Not comparable |
| MMLU-ProSource | 79.2% | — | Not comparable |
| HLESource | 49% | — | Not comparable |
| Artificial Analysis Intelligence IndexSource | 35.9% | 30.4% | Claude Sonnet 4.6 leads |
| AA-GPQA DiamondSource | 79.9% | 82.7% | o3 leads |
| AA-HLESource | 13.2% | 20.0% | o3 leads |
| AA-Omniscience IndexSource | -2.9% | -15.3% | Claude Sonnet 4.6 leads |
| AA-Omniscience AccuracySource | 38.0% | 38.4% | o3 leads |
| AA-Omniscience Hallucination RateSource | 65.9% | 87.1% | Claude Sonnet 4.6 leads |
MathClaude Sonnet 4.6 wins3 benchmarks
Multimodal3 benchmarks
Inst. Following1 benchmarks
| Benchmark | Claude Sonnet 4.6 | o3 | Result |
|---|---|---|---|
| AA-IFBenchSource | 41.2% | 71.4% | o3 leads |
Frequently Asked Questions (2)
Which is better, Claude Sonnet 4.6 or o3?
Claude Sonnet 4.6 is ahead on BenchLM's BenchAlign leaderboard, 65.07 to 47.89. The biggest single separator in this matchup is FrontierMath v2 (Tiers 1-3), where the scores are 32.400% and 18.685%.
Which is better for math, Claude Sonnet 4.6 or o3?
Claude Sonnet 4.6 has the edge for math in this comparison, averaging 26.4 versus 14.5. Inside this category, FrontierMath v2 (Tiers 1-3) is the benchmark that creates the most daylight between them.
Related Comparisons
Choose a model with this week’s evidence
Join 2,000+ readers for ranking moves, pricing changes, and the claims that still need proof.
One email each week. Unsubscribe anytime.