Compare AI models
| Specifications | ||
ParametersThe learned values that shape how a model responds. B means billion and T means trillion. For mixture-of-experts models, the total count is larger than the number active for each token. Parameter count alone does not measure quality. | 550B | 0.8B/2B/4B/9B |
Context windowHow much information the model can work with in one request, measured in tokens — pieces of words or other data. This includes your question, attachments, conversation and reply. Limits can vary by model variant and provider. | 1M | 256k |
| API pricingUSD per 1M tokens · lower wins | ||
Cheapest inputLowest input rate across every provider on OpenRouter, the lab included. The cheapest endpoint may run a quantised build or a shorter context — see "Available from" on the model page. | $0.40DeepInfra | — |
Cheapest outputLowest output rate across every provider on OpenRouter, the lab included. May come from a different provider than the cheapest input. | $1.76DeepInfra | — |
| $210KT | — | |
| Benchmarks | ||
LiveCodeBenchCompetitive coding — Coding problems published so recently the AI can't have seen them in training — a contamination-free test of raw programming skill. Higher is better. | 89% | 65.6% |
GPQA DiamondScience — Graduate-level science questions in biology, physics, and chemistry — hard enough that subject-matter PhDs score around 65%. Higher is better. | 87% | 81.7% |
IFBenchInstruction following — Tests whether the AI can follow detailed instructions and satisfy multiple constraints at once. Higher is better. | 82.94% | 64.5% |
MMLU-ProKnowledge — accuracy on the ten-option MMLU-Pro benchmark | 86.8% | 82.5% |
BullshitBench v2Nonsense detection — Given a confidently-worded but nonsensical prompt, does the AI spot that it makes no sense and push back — instead of playing along and inventing an answer? The score is how often it clearly called out the nonsense. Higher is better. | 49% | — |
SWE-Bench ProAgentic coding — Can the AI fix real bugs in real software? It's handed actual problems from open-source projects and has to write code that genuinely solves them. Higher is better. | 49.66% | — |
SWE-Bench VerifiedCoding — Real coding tasks pulled from open-source projects — the AI has to find and fix actual bugs. A human-checked version of the original SWE-Bench. Higher is better. | 71.9% | — |
SWE-Bench MultilingualMultilingual coding — Like SWE-Bench, but the coding problems span many programming languages, not just one. Tests how broadly the AI can code. Higher is better. | 67.7% | — |
Terminal-Bench 2.1Agentic terminal coding — Can the AI work in a command-line terminal — running commands and finishing technical setup tasks the way a developer would? Higher is better. | 56.4% | — |
IFEvalAgentic tasks — following verifiable instructions in generated responses | — | 91.5% |
τ²-bench · retailAgentic tasks — task success in the retail domain of the second-generation customer-service evaluation | 73.68% | — |
τ²-bench · airlineAgentic tasks — task success in the airline domain of the second-generation customer-service evaluation | 79.75% | — |
τ²-bench · telecomAgentic tasks — task success in the telecom domain of the second-generation customer-service evaluation | 83% | — |
BrowseCompWeb browsing — Can the AI browse the web and track down hard-to-find answers? Higher is better. | 44.4% | — |
Humanity's Last Exam · text only, no toolsMultidisciplinary reasoning — accuracy on the 2,158 text-only questions without tools | 29.24% | — |
Humanity's Last Exam · text only, with toolsMultidisciplinary reasoning — The text-only questions from Humanity's Last Exam, answered with access to tools such as search or code execution. The score is the percentage answered correctly. | 37.4% | — |
OSWorld-VerifiedAgentic computer use — Can the AI actually operate a computer — clicking, typing, and using real apps — to finish tasks on its own? Higher is better. | — | 41.8% |
AA Intelligence IndexOverall intelligence — Artificial Analysis composite intelligence index across evals | 48 | — |
CharXiv ReasoningChart reasoning — Can the AI read and reason about complex charts and figures, not just text? Higher is better. | — | 73% |
MMMU-ProMultimodal reasoning — A tougher version of MMMU — college-level questions that mix images, diagrams, and text together. Higher is better. | — | 70.1% |
MMMUMultimodal — Tests the AI on understanding images and text together across many college subjects. Higher is better. | — | 78.4% |
| Overview | ||
| Company | NVIDIA | Qwen |
| Release date | Jun 4 2026 | Mar 2 2026 |
| Access | Open Weight | Open Weight |
| Model details | View model | View model |
Other comparisons
Nemotron 3 UltravsClaude Fable 5.2Qwen3.5 SmallvsClaude Fable 5.2Nemotron 3 UltravsGPT-6.2 AstraQwen3.5 SmallvsGPT-6.2 AstraNemotron 3 UltravsGemini 4 NeonQwen3.5 SmallvsGemini 4 NeonNemotron 3 UltravsMuse EmberQwen3.5 SmallvsMuse EmberNemotron 3 UltravsGrok 4.7Qwen3.5 SmallvsGrok 4.7Nemotron 3 UltravsDeepSeek-V4.1-Flash-1017Qwen3.5 SmallvsDeepSeek-V4.1-Flash-1017Frequently asked questions
Nemotron 3 Ultra leads Qwen3.5 Small on 4 of the 4 benchmarks they both report (LiveCodeBench, GPQA Diamond, IFBench, MMLU-Pro). Qwen3.5 Small shipped 94 days before Nemotron 3 Ultra, so benchmark comparisons should account for the intervening progress.
Nemotron 3 Ultra has 550B parameters, while Qwen3.5 Small has 0.8B/2B/4B/9B. Context windows are 1M (Nemotron 3 Ultra) vs 256k (Qwen3.5 Small).
On LiveCodeBench, Nemotron 3 Ultra leads at 89% vs Qwen3.5 Small at 65.6%. On GPQA Diamond, Nemotron 3 Ultra leads at 87% vs Qwen3.5 Small at 81.7%. On IFBench, Nemotron 3 Ultra leads at 82.94% vs Qwen3.5 Small at 64.5%. On MMLU-Pro, Nemotron 3 Ultra leads at 86.8% vs Qwen3.5 Small at 82.5%.