Kimi K2 Instruct vs Nemotron 3 Ultra 550B A55B vs Qwen3.6 27B
Qwen3.6 27B comes out ahead, 71 to 67 and 55 on our weighted score, though Kimi K2 Instruct is 26% cheaper per token.
Price in rating: Excluded · Include price
Moonshot AI
Kimi K2 Instruct
55/100- ECI141.3
- Price$0.57 / $2.30
- Context131K
NVIDIA
Nemotron 3 Ultra 550B A55B
67/100- ECI147.0
- Price$0.55 / $2.40
- Context1M
- Our pick
Alibaba (Qwen)
Qwen3.6 27B
71/100- ECI146.8
- Price$0.60 / $3.60
- Context262K
Qwen3.6 27B is our pick
Qwen3.6 27B is the better all-round choice, scoring 71/100 against Nemotron 3 Ultra 550B A55B (67) and Kimi K2 Instruct (55). It leads on inputs & features. Nemotron 3 Ultra 550B A55B wins on context window. The score weighs capability 67%, inputs & features 20%, context window 13%.
- CapabilityNemotron 3 Ultra 550B A55BCapabilities Index (ECI): Nemotron 3 Ultra 550B A55B 147.0 · Qwen3.6 27B 146.8 · Kimi K2 Instruct 141.3
- Lowest priceKimi K2 InstructKimi K2 Instruct $1.00 · Nemotron 3 Ultra 550B A55B $1.01 · Qwen3.6 27B $1.35 per 1M tokens (3:1 blend)
- Longest contextNemotron 3 Ultra 550B A55BNemotron 3 Ultra 550B A55B 1,000,000 · Qwen3.6 27B 262,144 · Kimi K2 Instruct 131,072 tokens
- Widest inputsQwen3.6 27BKimi K2 Instruct: Text · Nemotron 3 Ultra 550B A55B: Text · Qwen3.6 27B: Text, Images, Video
- Self-hostingAll open weightsEvery model here can be downloaded and run on your own hardware
| Measure | Weight | Kimi K2 Instruct | Nemotron 3 Ultra 550B A55B | Qwen3.6 27B |
|---|---|---|---|---|
| CapabilityCapabilities Index (ECI) | 67% | 69 | 74 | 74 |
| Inputs & features | 20% | 25 | 45 | 80 |
| Context window | 13% | 32 | 64 | 43 |
| Overall | 100% | 55/100 | 67/100 | 71/100 |
Every spec in one table
Highlighted cells lead their row. Dashes mean the data is not published.
| Specification | |||
|---|---|---|---|
| Capability | |||
| Capabilities Index (ECI) | 141.3 | 147.0 (best) | 146.8 |
| ECI rank | #108 of 158 | #82 of 158 (best) | #84 of 158 |
| GPQA DiamondGraduate-level science questions | — | 85.4% | 85.9% (best) |
| FrontierMath Tiers 1–3Research-level mathematics | — | — | 35.1% |
| OTIS Mock AIME 2024–2025Competition mathematics | — | 86.7% | 91.1% (best) |
| Price per million tokens | |||
| Input | $0.57 | $0.55 (best) | $0.60 |
| Output | $2.30 (best) | $2.40 | $3.60 |
| Cached input | — | — | — |
| Blended (3:1) | $1.00 (best) | $1.01 | $1.35 |
| Long-context rate | Same rate | Same rate | Same rate |
| Price source | Median of 11 providers | Median of 20 providers | Official Alibaba API |
| Limits | |||
| Context window | 131,072 tokens | 1,000,000 tokens (best) | 262,144 tokens |
| Max output | 32,768 tokens | 128,000 tokens (best) | 65,536 tokens |
| Inputs and features | |||
| Text | Yes | Yes | Yes |
| Images | No | No | Yes |
| PDFs | No | No | No |
| Audio | No | No | No |
| Video | No | No | Yes |
| Reasoning | No | Yesmedium · high | Yes |
| Tool calling | Yes | Yes | Yes |
| Structured output | No | Yes | Yes |
| Availability | |||
| Weights | OpenModified MIT | Open | Open |
| API model ID | — | nvidia/nemotron-3-ultra-550b-a55b | qwen3.6-27b |
| API providers | 11 | 26 | 29 (best) |
| Released | Jul 11, 2025 | Jun 4, 2026 | Apr 22, 2026 |
| Knowledge cutoff | — | Sep 2025 | — |
What would a month cost?
Enter your expected volume in millions of tokens. List prices only; caching and batch discounts would lower these.
Kimi K2 Instruct$10.30
Nemotron 3 Ultra 550B A55B$10.30
Qwen3.6 27B$13.20
Which should you choose?
Which is better: Kimi K2 Instruct, Nemotron 3 Ultra 550B A55B or Qwen3.6 27B?
Qwen3.6 27B is the better all-round choice, scoring 71/100 against Nemotron 3 Ultra 550B A55B (67) and Kimi K2 Instruct (55). It leads on inputs & features. Nemotron 3 Ultra 550B A55B wins on context window. The score weighs capability 67%, inputs & features 20%, context window 13%.
Which is cheaper, Kimi K2 Instruct, Nemotron 3 Ultra 550B A55B or Qwen3.6 27B?
Kimi K2 Instruct is cheaper at $0.57 input / $2.30 output per million tokens (median across 11 API providers). Nemotron 3 Ultra 550B A55B costs $0.55 input / $2.40 output per million tokens (median across 20 API providers; free on Nvidia); Qwen3.6 27B costs $0.60 input / $3.60 output per million tokens (official Alibaba API price). At a typical mix of three input tokens to one output token, that is $1.00 per million tokens for Kimi K2 Instruct versus $1.01 for Nemotron 3 Ultra 550B A55B (1× as much) and $1.35 for Qwen3.6 27B (1.3× as much).
Which scores higher on benchmarks?
Nemotron 3 Ultra 550B A55B scores higher on the Capabilities Index (ECI): Nemotron 3 Ultra 550B A55B 147.0 (#82 of 158), Qwen3.6 27B 146.8 (#84 of 158) and Kimi K2 Instruct 141.3 (#108 of 158). The confidence ranges of the top two overlap (145.0–149.1 vs 144.7–148.9), so treat the gap as small.
Which is better for coding?
There are no published SWE-bench Verified results for Kimi K2 Instruct, Nemotron 3 Ultra 550B A55B and Qwen3.6 27B yet, so there is no like-for-like coding score. On overall capability, Nemotron 3 Ultra 550B A55B leads, which tends to carry over to coding, but test on your own codebase. All three support tool calling for agent workflows.
Which has the bigger context window?
Nemotron 3 Ultra 550B A55B has the largest context window at 1,000,000 tokens, against 262,144 for Qwen3.6 27B and 131,072 for Kimi K2 Instruct. Maximum output per response: Kimi K2 Instruct up to 32,768, Nemotron 3 Ultra 550B A55B up to 128,000, Qwen3.6 27B up to 65,536 tokens.
Which can read images, PDFs, audio or video?
Kimi K2 Instruct accepts text; Nemotron 3 Ultra 550B A55B accepts text; Qwen3.6 27B accepts text, images and video. Qwen3.6 27B handles the widest range of inputs.
Are any of these open source?
Yes, all three publish their weights (Modified MIT), so you can self-host them.
Which is newer?
Nemotron 3 Ultra 550B A55B is the newest, released Jun 4, 2026. Qwen3.6 27B came out Apr 22, 2026; Kimi K2 Instruct came out Jul 11, 2025. Knowledge cutoff: Nemotron 3 Ultra 550B A55B Sep 2025.
How do you decide the winner?
This rating uses capability (67%, independent benchmark results); inputs & features (20%, image, PDF, audio and video input, tool calling, structured output and reasoning); context window (13%, maximum tokens per request, log scale). Dimensions missing for any model are dropped and the remaining weights rescaled, so every model is judged on the same evidence. Specs and prices come from public model listings and the labs’ own API pages; capability scores come from independent benchmark runs. Data updated Nov 18, 2026.