Compare
Four models, one page
Pick the models you are actually choosing between. Every row is scaled within itself, so a bar means something next to its neighbours and nothing across rows.
| Specification | Kimi K2.5Moonshot AI | Claude Opus 5Anthropic |
|---|---|---|
| Released | Jan 27, 2026 | Jul 24, 2026 |
| Context window | 262K | 1M |
| Max output | 262K | 128K |
| Input / 1M | $0.57 | $5 |
| Output / 1M | $2.85 | $25 |
| Cost per task | $0.26 | $2.25 |
| Arena Elo | 1,272 | 1,393 |
| Serving providers | 10 | 5 |
| Parameters | 1058.6B | Undisclosed |
| Licence | other | Proprietary |
| Capabilities |
|
|
Metrics side by side
Each row is scaled to the largest value in that row — bars compare within a row, never across rows
- Kimi K2.5
- Claude Opus 5
Intelligence Index
Kimi K2.536.0Claude Opus 563.1Coding Index
Kimi K2.546.8Claude Opus 578.0Agentic Index
Kimi K2.521.7Claude Opus 559.2Output speed
Kimi K2.573 t/sClaude Opus 587 t/sContext window
Kimi K2.5262KClaude Opus 51MLatency · lower is better
Kimi K2.5849msClaude Opus 51.12sBlended price / 1M · lower is better
Kimi K2.5$1.14Claude Opus 5$10Cost per task · lower is better
Kimi K2.5$0.26Claude Opus 5$2.25
Rows marked “lower is better” still draw a longer bar for a larger number — read the value, not just the length. Arena Elo is in the specification table above instead: it has no meaningful zero, so a bar would flatten the gaps.
View as table
| Metric | Kimi K2.5 | Claude Opus 5 |
|---|---|---|
| Intelligence Index | 36.0 | 63.1 |
| Coding Index | 46.8 | 78.0 |
| Agentic Index | 21.7 | 59.2 |
| Output speed | 73 t/s | 87 t/s |
| Context window | 262K | 1M |
| Latency · lower is better | 849ms | 1.12s |
| Blended price / 1M · lower is better | $1.14 | $10 |
| Cost per task · lower is better | $0.26 | $2.25 |
Evaluation scores
Percentage correct on a common 0–100% scale
- Kimi K2.5
- Claude Opus 5
GPQA Diamond
Kimi K2.587.9%Claude Opus 593.2%Humanity's Last Exam
Kimi K2.530.7%Claude Opus 554.9%SciCode
Kimi K2.549.0%Claude Opus 555.7%τ²-bench
Kimi K2.595.9%Claude Opus 542.1%Terminal-Bench Hard
Kimi K2.534.8%Claude Opus 5not measuredAA-LCR long context
Kimi K2.573.0%Claude Opus 575.7%
A missing bar means that evaluation was not run for that model — it is not a zero.
View as table
| Evaluation | Kimi K2.5 | Claude Opus 5 |
|---|---|---|
| GPQA Diamond | 87.9% | 93.2% |
| Humanity's Last Exam | 30.7% | 54.9% |
| SciCode | 49.0% | 55.7% |
| τ²-bench | 95.9% | 42.1% |
| Terminal-Bench Hard | 34.8% | — |
| LiveCodeBench | — | — |
| AA-LCR long context | 73.0% | 75.7% |
| AIME 2025 | — | — |
Estimated from list pricing: 50K input tokens plus 80K output tokens for reasoning models (25K for non-reasoning).