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. | 30B | 4B/30B/235B |
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. | $0.00Nvidia | — |
Cheapest outputLowest output rate across every provider on OpenRouter, the lab included. May come from a different provider than the cheapest input. | $0.00Nvidia | — |
| $11.00Darkbloom · int4 | — | |
| Benchmarks | ||
Humanity's Last Exam · text only, no toolsMultidisciplinary reasoning — accuracy on the 2,158 text-only questions without tools | 11.72% | 18.2% |
GPQA DiamondScience — Graduate-level science questions in biology, physics, and chemistry — hard enough that subject-matter PhDs score around 65%. Higher is better. | 75.44% | 81.1% |
MMLU-ProKnowledge — accuracy on the ten-option MMLU-Pro benchmark | 81.94% | 84.4% |
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. | 51.56% | — |
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. | 39.33% | — |
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. | — | 74.1% |
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. | 24.58% | — |
IFEvalAgentic tasks — following verifiable instructions in generated responses | — | 87.8% |
BrowseCompWeb browsing — Can the AI browse the web and track down hard-to-find answers? Higher is better. | 36.97% | — |
AIME 2025Reasoning — single-answer accuracy on the 2025 invitational mathematics exam without tools | — | 92.3% |
| Overview | ||
| Company | NVIDIA | Qwen |
| Release date | Aug 11 2026 | Jul 21 2025 |
| Access | Open Weight | Open Weight |
| Model details | View model | View model |
Other comparisons
Nemotron 3.5 LightningvsClaude Fable 5.2Qwen3-2507vsClaude Fable 5.2Nemotron 3.5 LightningvsGPT-6.2 AstraQwen3-2507vsGPT-6.2 AstraNemotron 3.5 LightningvsGemini 4 NeonQwen3-2507vsGemini 4 NeonNemotron 3.5 LightningvsMuse EmberQwen3-2507vsMuse EmberNemotron 3.5 LightningvsGrok 4.7Qwen3-2507vsGrok 4.7Nemotron 3.5 LightningvsDeepSeek-V4.1-Flash-1017Qwen3-2507vsDeepSeek-V4.1-Flash-1017Frequently asked questions
Qwen3-2507 leads Nemotron 3.5 Lightning on 3 of the 3 benchmarks they both report (Humanity's Last Exam, GPQA Diamond, MMLU-Pro). Qwen3-2507 shipped 386 days before Nemotron 3.5 Lightning, so benchmark comparisons should account for the intervening progress.
Nemotron 3.5 Lightning has 30B parameters, while Qwen3-2507 has 4B/30B/235B. Context windows are 1M (Nemotron 3.5 Lightning) vs 256k (Qwen3-2507).
On Humanity's Last Exam · text only, no tools, Qwen3-2507 leads at 18.2% vs Nemotron 3.5 Lightning at 11.72%. On GPQA Diamond, Qwen3-2507 leads at 81.1% vs Nemotron 3.5 Lightning at 75.44%. On MMLU-Pro, Qwen3-2507 leads at 84.4% vs Nemotron 3.5 Lightning at 81.94%.