Model comparison
Kimi K2.5 vs Nemotron 3 Ultra
Head-to-head evidence from 30 shared benchmark results across 7 categories. Overall scores shown here use the public BenchAlign v5 ranking lane.
Verified leaderboard positions: Kimi K2.5 #22; Nemotron 3 Ultra #24
BenchAlign evidence: Kimi K2.5 supported; Nemotron 3 Ultra not scored. Intervals and evidence labels describe ranking uncertainty, not a guarantee for a specific workload.
Evidence parity. Kimi K2.5 and Nemotron 3 Ultra share 30 comparable benchmark results. 6 of 8 categories are comparable. 34 results are unique to Kimi K2.5; 9 to Nemotron 3 Ultra.
Updated July 16, 2026- Shared results
- 30
- Kimi K2.5 only
- 34
- Nemotron 3 Ultra only
- 9
- Comparable categories
- 6 / 8
Pick Kimi K2.5 if you want the stronger benchmark profile. Nemotron 3 Ultra only becomes the better choice if reasoning is the priority or you want the cheaper token bill.
Confidence note. This is a partial-evidence comparison with 30 shared benchmark results across 7 evidence categories; 6 of 8 categories currently have scoreable aggregates for both models. Treat the verdict as directional until coverage is more balanced.
Why this result
Kimi K2.5 finishes one point ahead on BenchLM's provisional leaderboard, 61 to 60. That is enough to call, but not enough to treat as a blowout. This matchup comes down to a few meaningful edges rather than one model dominating the board.
Kimi K2.5's sharpest advantage is in instruction following, where it averages 93.9 against 81.7. The single biggest benchmark swing on the page is BrowseComp, 60.6% to 44.4%. Nemotron 3 Ultra does hit back in reasoning, so the answer changes if that is the part of the workload you care about most.
Kimi K2.5 is also the more expensive model on tokens at $0.60 input / $3.00 output per 1M tokens, versus $0.00 input / $0.00 output per 1M tokens for Nemotron 3 Ultra. That is roughly Infinityx on output cost alone. Nemotron 3 Ultra is the reasoning model in the pair, while Kimi K2.5 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. Nemotron 3 Ultra gives you the larger context window at 1M, compared with 256K for Kimi K2.5.
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 | Kimi K2.5 | Δ | Nemotron 3 Ultra |
|---|---|---|---|
| Inst. Following | Kimi K2.593.9 | Margin← 12.2 | Nemotron 3 Ultra81.7 |
| Agentic | Kimi K2.555.0 | Margin← 3.7 | Nemotron 3 Ultra51.3 |
| Knowledge | Kimi K2.557.2 | Margin← 3.0 | Nemotron 3 Ultra54.2 |
| Coding | Kimi K2.559.4 | Margin← 1.1 | Nemotron 3 Ultra58.3 |
| Reasoning | Kimi K2.561.0 | Margin→ 0.9 | Nemotron 3 Ultra61.9 |
| Multilingual | Kimi K2.582.3 | Margin→ 0.7 | Nemotron 3 Ultra83.0 |
| Math | Kimi K2.560.6 | MarginNo overlap | Nemotron 3 UltraNot measured |
| Multimodal | Kimi K2.578.5 | MarginNo overlap | Nemotron 3 UltraNot measured |
Decisive benchmark drivers
The largest measured benchmark gaps in this matchup, with exact reported values.
More
- Source ↗
BrowseComp
AgenticA 60.6%B 44.4%Winner: Kimi K2.5Δ 16.2BrowseComp: Kimi K2.5 scored 60.6%; Nemotron 3 Ultra scored 44.4%. Kimi K2.5 wins this benchmark. - Source ↗
Terminal-Bench 2.0
AgenticA 50.8%B 56.4%Winner: Nemotron 3 UltraΔ 5.6Terminal-Bench 2.0: Kimi K2.5 scored 50.8%; Nemotron 3 Ultra scored 56.4%. Nemotron 3 Ultra wins this benchmark. - Source ↗
SWE-bench Verified
CodingA 76.8%B 71.9%Winner: Kimi K2.5Δ 4.9SWE-bench Verified: Kimi K2.5 scored 76.8%; Nemotron 3 Ultra scored 71.9%. Kimi K2.5 wins this benchmark. - Source ↗
SciCode
CodingA 48.7%B 44.6%Winner: Kimi K2.5Δ 4.1SciCode: Kimi K2.5 scored 48.7%; Nemotron 3 Ultra scored 44.6%. Kimi K2.5 wins this benchmark. - Source ↗
HLE
KnowledgeA 30.1%B 26.7%Winner: Kimi K2.5Δ 3.4HLE: Kimi K2.5 scored 30.1%; Nemotron 3 Ultra scored 26.7%. Kimi K2.5 wins this benchmark.
Operational comparison
Runtime and commercial metrics are compared only when both models have a complete sourced value.
| Metric | Kimi K2.5 | Nemotron 3 Ultra | Comparison |
|---|---|---|---|
| Input / output priceUSD per 1M tokens | Kimi K2.5$0.6 input / $3 output | Nemotron 3 Ultra$0 input / $0 output | Nemotron 3 Ultra has the lower combined listed price. |
| Generation speedtokens per second | Kimi K2.545 tok/s | Nemotron 3 UltraNot available | A complete speed comparison is not available. |
| First-answer latencyseconds to first token | Kimi K2.52.38 s | Nemotron 3 UltraNot available | A complete latency comparison is not available. |
| Context windowmaximum listed tokens | Kimi K2.5256K | Nemotron 3 Ultra1M | Nemotron 3 Ultra lists the larger context window. |
Benchmark Deep Dive
AgenticKimi K2.5 wins24 benchmarks
| Benchmark | Kimi K2.5 | Nemotron 3 Ultra | Result |
|---|---|---|---|
| Terminal-Bench 2.0Source | 50.8% | 56.4% | Nemotron 3 Ultra leads |
| BrowseCompSource | 60.6% | 44.4% | Kimi K2.5 leads |
| Claw-EvalSource | 52.3% | — | Not comparable |
| QwenClawBenchSource | 54.3% | — | Not comparable |
| τ³-bench resultsSource | 65.7% | 70.9% | Nemotron 3 Ultra leads |
| DeepSearchQASource | 77.1% | — | Not comparable |
| DeepPlanningSource | 14.4% | — | Not comparable |
| ToolathlonSource | 27.8% | — | Not comparable |
| MCP AtlasSource | 29.5% | — | Not comparable |
| MCP-TasksSource | 59.1% | — | Not comparable |
| WideResearchSource | 72.7% | — | Not comparable |
| τ²-bench resultsSource | 95.9% | 83.3% | Kimi K2.5 leads |
| APEX-Agents-AASource | 11.5% | — | Not comparable |
| Gert LabsSource | 45.88% | — | Not comparable |
| ResearchClawBenchSource | 14.0% | — | Not comparable |
| JobBenchSource | 8.7% | — | Not comparable |
| AA Agentic IndexSource | 21.7% | 27.4% | Nemotron 3 Ultra leads |
| GDPval-AASource | 25.4% | 33.2% | Nemotron 3 Ultra leads |
| GDPval-AASource | 1009 | 1164 | Nemotron 3 Ultra leads |
| PinchBenchSource | — | 90.0% | Not comparable |
| HLE w/ toolsSource | — | 37.4% | Not comparable |
| AA BriefcaseSource | — | 870 | Not comparable |
| AA EnterpriseOps-GymSource | — | 28.9% | Not comparable |
| AA Harvey LABSource | — | 3.3% | Not comparable |
CodingKimi K2.5 wins12 benchmarks
| Benchmark | Kimi K2.5 | Nemotron 3 Ultra | Result |
|---|---|---|---|
| SWE-bench VerifiedSource | 76.8% | 71.9% | Kimi K2.5 leads |
| SWE-bench Verified*Source | 70.8% | — | Not comparable |
| LiveCodeBench v6Source | 85.0% | 89.0% | Nemotron 3 Ultra leads |
| SWE-bench ProSource | 50.7% | — | Not comparable |
| SWE MultilingualSource | 73% | 67.7% | Kimi K2.5 leads |
| SWE-RebenchSource | 58.5% | — | Not comparable |
| React Native EvalsSource | 77.2% | — | Not comparable |
| SciCodeSource | 48.7% | 44.6% | Kimi K2.5 leads |
| Terminal-Bench HardSource | 34.8% | 36.4% | Nemotron 3 Ultra leads |
| AA-SciCodeSource | 49.0% | 39.9% | Kimi K2.5 leads |
| AA Coding IndexSource | 46.8% | 49.3% | Nemotron 3 Ultra leads |
| Terminal-Bench 2.0Source | — | 56.4% | Not comparable |
ReasoningNemotron 3 Ultra wins3 benchmarks
KnowledgeKimi K2.5 wins14 benchmarks
| Benchmark | Kimi K2.5 | Nemotron 3 Ultra | Result |
|---|---|---|---|
| GPQASource | 87.6% | 87% | Kimi K2.5 leads |
| GPQA-DSource | 87.6% | 87.0% | Kimi K2.5 leads |
| SuperGPQASource | 69.2% | — | Not comparable |
| MMLU-ProSource | 87.1% | 86.8% | Kimi K2.5 leads |
| MMLU-Pro (Arcee)Source | 87.1% | — | Not comparable |
| HLESource | 30.1% | 26.7% | Kimi K2.5 leads |
| Artificial Analysis Intelligence IndexSource | 35.4% | 37.8% | Nemotron 3 Ultra leads |
| AA-GPQA DiamondSource | 87.9% | 86.7% | Kimi K2.5 leads |
| AA-HLESource | 29.4% | 26.6% | Kimi K2.5 leads |
| AA-Omniscience IndexSource | -8.1% | -0.8% | Nemotron 3 Ultra leads |
| AA-Omniscience AccuracySource | 34.3% | 21.6% | Kimi K2.5 leads |
| AA-Omniscience Hallucination RateSource | 64.6% | 28.5% | Nemotron 3 Ultra leads |
| HLE w/o toolsSource | — | 26.7% | Not comparable |
| AA Openness IndexSource | — | 83.3% | Not comparable |
Math9 benchmarks
| Benchmark | Kimi K2.5 | Nemotron 3 Ultra | Result |
|---|---|---|---|
| AIME 2025Source | 96.1% | — | Not comparable |
| AIME26Source | 95.8% | — | Not comparable |
| AIME25 (Arcee)Source | 96.3% | — | Not comparable |
| HMMT Feb 2025Source | 95.4% | — | Not comparable |
| HMMT Nov 2025Source | 91.1% | — | Not comparable |
| HMMT Feb 2026Source | 87.1% | — | Not comparable |
| MMAnswerBenchSource | 81.8% | — | Not comparable |
| FrontierMath v2 (Tiers 1-3)Source | 27.900% | — | Not comparable |
| FrontierMath v2 (Tier 4)Source | 4.200% | — | Not comparable |
MultilingualNemotron 3 Ultra wins2 benchmarks
Multimodal6 benchmarks
Frequently Asked Questions (7)
Which is better, Kimi K2.5 or Nemotron 3 Ultra?
Kimi K2.5 is ahead on BenchLM's provisional leaderboard, 61 to 60. The biggest single separator in this matchup is BrowseComp, where the scores are 60.6% and 44.4%.
Which is better for knowledge tasks, Kimi K2.5 or Nemotron 3 Ultra?
Kimi K2.5 has the edge for knowledge tasks in this comparison, averaging 57.2 versus 54.2. Inside this category, AA-Omniscience Hallucination Rate is the benchmark that creates the most daylight between them.
Which is better for coding, Kimi K2.5 or Nemotron 3 Ultra?
Kimi K2.5 has the edge for coding in this comparison, averaging 59.4 versus 58.3. Inside this category, AA-SciCode is the benchmark that creates the most daylight between them.
Which is better for reasoning, Kimi K2.5 or Nemotron 3 Ultra?
Nemotron 3 Ultra has the edge for reasoning in this comparison, averaging 61.9 versus 61. Inside this category, AA-LCR is the benchmark that creates the most daylight between them.
Which is better for agentic tasks, Kimi K2.5 or Nemotron 3 Ultra?
Kimi K2.5 has the edge for agentic tasks in this comparison, averaging 55 versus 51.3. Inside this category, GDPval-AA is the benchmark that creates the most daylight between them.
Which is better for instruction following, Kimi K2.5 or Nemotron 3 Ultra?
Kimi K2.5 has the edge for instruction following in this comparison, averaging 93.9 versus 81.7. Inside this category, AA-IFBench is the benchmark that creates the most daylight between them.
Which is better for multilingual tasks, Kimi K2.5 or Nemotron 3 Ultra?
Nemotron 3 Ultra has the edge for multilingual tasks in this comparison, averaging 83 versus 82.3. Inside this category, MMLU-ProX is the benchmark that creates the most daylight between them.
Self-host vs API cost
Estimates at 50,000 req/day · 1000 tokens/req average.
Related Comparisons
The AI models change fast. We track them for you.
A weekly brief for engineers and researchers covering new models, ranking shifts, and pricing changes.
Free. No spam. Unsubscribe anytime.