Kimi K3 vs Nemotron 3.5 Super VL
Wins 4 of 5 areas
Coding · Agents · Reasoning · Long documents
Wins 0 of 5 areas
—
Kimi K3 is the stronger all-rounder.Nemotron 3.5 Super VL is cheaper.
Scores updated · 29 tests both models report · How we compare
Where each one wins
Tests won in each of the five areas where both have results. Each piece is one test, so a longer bar means more evidence; grey means the two scored within a point of each other.
- AgentsCarrying out multi-step tasks on its own40Kimi K34 of 4 tests
- ReasoningHard problems that need careful thinking30Kimi K33 of 3 tests
- CodingWriting and fixing software20Kimi K32 of 2 tests
- Long documentsFinding answers in very long texts20Kimi K32 of 2 tests
- FactsGetting facts right instead of making them up11Even1 each
What it costs
Prices per million tokens, roughly 750,000 words. The bars show the cost of a million tokens read plus a million written.
Nemotron 3.5 Super VL costs 96% less for the same work.
The biggest differences
The tests each model wins by the widest margin, up to three each. Scores are out of 100.
Where Kimi K3 pulls ahead
- Abstract visual puzzles that people can solveARC-AGI-2+58.9points ahead
- Command-line tasks in a real terminalTerminal-Bench 2.1+24.3points ahead
- Answers hard knowledge questions correctlyAA-Omniscience · Accuracy+24.3points ahead
Where Nemotron 3.5 Super VL pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+17.4points ahead
Every test, side by side
All 29 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingKimi K3
- Terminal-Bench 2.1Kimi K3 by 24.38560.7+24.3
- SciCodeKimi K3 by 11.759.547.8+11.7
AgentsKimi K3
- GDPValKimi K3 by 18.651.733.1+18.6
- MCP AtlasKimi K3 by 16.584.267.7+16.5
- Terminal-Bench 4.0Kimi K3 by 12.612.60+12.6
- BrowseCompKimi K3 by 12.391.278.9+12.3
ReasoningKimi K3
- ARC-AGI-2Kimi K3 by 58.960.41.5+58.9
- Humanity's Last ExamKimi K3 by 16.246.930.7+16.2
- CritPtKimi K3 by 12.823.410.6+12.8
FactsEven
- AA-Omniscience · AccuracyKimi K3 by 24.347.623.3+24.3
- AA-Omniscience · Non-hallucinationNemotron 3.5 Super VL by 17.446.864.2+17.4
Long documentsKimi K3
- AA-LCRKimi K3 by 19.788.769+19.7
- LongBench v2Kimi K3 by 6.96962.1+6.9
Other results16 tests, not counted
Tests outside the eight areas. They are not counted above: several are summary scores built from other tests, or the same test under another name.
- ARC-AGI-1Kimi K3 by 71.894.522.7+71.8
- DeepSWE 1.1Kimi K3 by 44.36924.7+44.3
- SWEBench Pro v2 (Hard)Kimi K3 by 44.188.244.1+44.1
- GDPval-AA 2.1Kimi K3 by 381 rating points15431162+381 rating
- ScreenSpot Pro originalKimi K3 by 30.373.843.5+30.3
- CritPt v2 HardKimi K3 by 2740.513.5+27
- AA-OmniscienceKimi K3 by 23.919.7-4.2+23.9
- HLE-DiamondKimi K3 by 18.126.98.8+18.1
- DeepSearchQA (F1)Kimi K3 by 17.39577.7+17.3
- AA IntelligenceKimi K3 by 17.143.626.5+17.1
- Vals Finance IndexKimi K3 by 175841+17
- Chartography originalKimi K3 by 16.926.69.7+16.9
- JobBenchKimi K3 by 16.452.936.5+16.4
- GDP (Surge AI)Kimi K3 by 14228+14
- Terminal-Bench 4.0Kimi K3 by 12.612.60+12.6
- Terminal-Bench-Science 0.1Kimi K3 by 8.18.10+8.1
Questions people ask
Which is better, Kimi K3 or Nemotron 3.5 Super VL?
Kimi K3 wins four of the five areas where both have results: coding, agents, reasoning and long documents. Nemotron 3.5 Super VL wins none, but costs 96% less. They are level on facts.
Which is better for coding?
Kimi K3. It wins 2 of the 2 coding tests both models report; Nemotron 3.5 Super VL wins none.
Which is cheaper?
Kimi K3 costs $3.00 per million input tokens and $15.00 per million output tokens; Nemotron 3.5 Super VL costs $0.15 and $0.50. That makes Nemotron 3.5 Super VL about 96% cheaper for the same work.
How do you compare the two?
We use the 29 benchmark tests both models have published scores on. The verdict counts the 13 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 16 are listed but not counted, because several are summary scores or repeat a test. Each score is the one shown on the model's own page.