Kimi K2.6 vs Nemotron 3 Super 120B A12B
Wins 7 of 7 areas
Coding · Agents · Reasoning · Facts · Math · Long documents · Following instructions
Wins 0 of 7 areas
—
Kimi K2.6 is the stronger all-rounder.Nemotron 3 Super 120B A12B is cheaper.
Scores updated · 28 tests both models report · How we compare
Where each one wins
Tests won in each of the seven 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.
- CodingWriting and fixing software50Kimi K2.65 of 5 tests
- AgentsCarrying out multi-step tasks on its own40Kimi K2.64 of 5 tests · 1 tie
- ReasoningHard problems that need careful thinking30Kimi K2.63 of 3 tests
- FactsGetting facts right instead of making them up20Kimi K2.62 of 2 tests
- MathCompetition and research-level math20Kimi K2.62 of 2 tests
- Long documentsFinding answers in very long texts10Kimi K2.61 of 1 test
- Following instructionsDoing exactly what it is asked10Kimi K2.61 of 1 test
Long documents and following instructions rest on a single test 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 Super 120B A12B costs 76% 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 K2.6 pulls ahead
- Finds hard-to-locate facts by browsing the webBrowseComp+55points ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+46.5points ahead
- Fixes real GitHub issues in many programming languagesSWE-bench Multilingual+30.9points ahead
Where Nemotron 3 Super 120B A12B pulls ahead
No clear win on a test scored out of 100.
Every test, side by side
All 28 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingKimi K2.6
- SWE-bench MultilingualKimi K2.6 by 30.976.745.8+30.9
- Terminal-Bench 2.1Kimi K2.6 by 27.365.938.6+27.3
- SWE-bench VerifiedKimi K2.6 by 26.580.253.7+26.5
- SciCodeKimi K2.6 by 15.351.536.2+15.3
- Terminal-Bench HardKimi K2.6 by 15.143.928.8+15.1
AgentsKimi K2.6
- BrowseCompKimi K2.6 by 5586.331.3+55
- GDPValKimi K2.6 by 27270+27
- AA ApexAgentsKimi K2.6 by 26.728.51.8+26.7
- τ-Bench V3 · BankingKimi K2.6 by 1323.310.3+13
- Terminal-Bench 4.0tie0.50tie
ReasoningKimi K2.6
- Humanity's Last ExamKimi K2.6 by 16.737.520.8+16.7
- GPQA DiamondKimi K2.6 by 11.191.180+11.1
- CritPtKimi K2.6 by 4.983.1+4.9
FactsKimi K2.6
- AA-Omniscience · Non-hallucinationKimi K2.6 by 46.559.513+46.5
- AA-Omniscience · AccuracyKimi K2.6 by 8.332.624.3+8.3
MathKimi K2.6
- HMMT Feb. 2026Kimi K2.6 by 7.892.784.8+7.8
- AIME 2026Kimi K2.6 by 4.796.491.7+4.7
Other results9 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.
- AA-OmniscienceKimi K2.6 by 46.85.3-41.5+46.8
- Terminal-Bench 2.0Kimi K2.6 by 35.766.731+35.7
- τ²-Bench Telecom (AA run)Kimi K2.6 by 28.195.967.8+28.1
- Artificial Analysis Coding IndexKimi K2.6 by 24.161.837.7+24.1
- AA Agentic IndexKimi K2.6 by 1822.14.1+18
- AA IntelligenceKimi K2.6 by 14.22712.8+14.2
- SciCode (subtask)Kimi K2.6 by 9.75242.3+9.7
- LiveCodeBenchKimi K2.6 by 8.489.681.2+8.4
- MMLU-ProKimi K2.6 by 3.487.183.7+3.4
Questions people ask
Which is better, Kimi K2.6 or Nemotron 3 Super 120B A12B?
Kimi K2.6 wins all seven areas where both have results: coding, agents, reasoning, facts, math, long documents and following instructions. Nemotron 3 Super 120B A12B wins none, but costs 76% less.
Which is better for coding?
Kimi K2.6. It wins 5 of the 5 coding tests both models report; Nemotron 3 Super 120B A12B wins none.
Which is cheaper?
Kimi K2.6 costs $0.95 per million input tokens and $4.00 per million output tokens; Nemotron 3 Super 120B A12B costs $0.30 and $0.90. That makes Nemotron 3 Super 120B A12B about 76% cheaper for the same work.
How do you compare the two?
We use the 28 benchmark tests both models have published scores on. The verdict counts the 19 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 9 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.