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. | 4B | — |
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. | 256k | 1M |
| 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. | — | $0.10Azure |
Cheapest outputLowest output rate across every provider on OpenRouter, the lab included. May come from a different provider than the cheapest input. | — | $0.40Azure |
| — | $40.00Azure | |
| Benchmarks | ||
GPQA DiamondScience — Graduate-level science questions in biology, physics, and chemistry — hard enough that subject-matter PhDs score around 65%. Higher is better. | 72.41% | 50.3% |
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. | 59.5% | — |
Aider Polyglot (diff)Coding — editing code in six languages using the diff edit format | — | 6.2% |
HumanEvalFunction synthesis — 164 small Python problems: the AI is given a function's description and has to write the working function. This was the coding benchmark of the GPT-3.5 and GPT-4 era, before the field moved to fixing real bugs in real repositories. Higher is better. | — | 87% |
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. | 18.1% | — |
IFEvalAgentic tasks — following verifiable instructions in generated responses | — | 74.5% |
MATH-500Reasoning — accuracy on the 500-problem MATH subset | — | 62.3% |
DROPReasoning — three-shot token F1 on questions requiring reasoning over paragraphs | — | 82.2 |
AIME 2024Reasoning — single-answer accuracy on the 2024 invitational mathematics exam without tools | — | 29.4% |
IFBenchInstruction following — Tests whether the AI can follow detailed instructions and satisfy multiple constraints at once. Higher is better. | 77.94% | — |
SimpleQAKnowledge — accuracy on short, fact-seeking questions in the original SimpleQA set | — | 7.6% |
MMMLUKnowledge — average accuracy on multilingual translations of MMLU | — | 66.9% |
MMLUGeneral knowledge — A 57-subject multiple-choice exam — history, law, medicine, maths — that was the standard measure of how much a model knows from 2020 until roughly 2024, when frontier scores crowded into the high 80s and labs moved on to harder tests. The scores here were published years apart under different testing setups, so read them as a historical record rather than a like-for-like ranking. Higher is better. | — | 80.1% |
MGSMMultilingual math — solving grade-school word problems translated into multiple languages | — | 73% |
CharXiv ReasoningChart reasoning — Can the AI read and reason about complex charts and figures, not just text? Higher is better. | — | 40.5% |
MathVistaMultimodal reasoning — mathematical reasoning about visual information | — | 56.2% |
MMMUMultimodal — Tests the AI on understanding images and text together across many college subjects. Higher is better. | — | 55.4% |
GraphWalks BFS (<128K)Long context — breadth-first traversal accuracy for inputs shorter than 128K tokens | — | 25% |
GraphWalks parents (<128K)Long context — parent-node retrieval accuracy for inputs shorter than 128K tokens | — | 9.4% |
OpenAI MRCR (2 needles) · 128KLong context — retrieval of two matching turns in the original MRCR evaluation | — | 36.6% |
OpenAI MRCR (2 needles) · 1MLong context — retrieval of two matching turns in the original MRCR evaluation | — | 12% |
| Overview | ||
| Company | NVIDIA | OpenAI |
| Release date | Oct 16 2026 | Apr 14 2025 |
| Access | Open Weight | Closed |
| Model details | View model | View model |
Other comparisons
Nemotron Labs Dory 1.0 4BvsClaude Fable 5.2GPT-4.1 nanovsClaude Fable 5.2Nemotron Labs Dory 1.0 4BvsGemini 4 CarbonGPT-4.1 nanovsGemini 4 CarbonNemotron Labs Dory 1.0 4BvsMuse EmberGPT-4.1 nanovsMuse EmberNemotron Labs Dory 1.0 4BvsGrok 4.7GPT-4.1 nanovsGrok 4.7Nemotron Labs Dory 1.0 4BvsDeepSeek-V4.1-Flash-1017GPT-4.1 nanovsDeepSeek-V4.1-Flash-1017Nemotron Labs Dory 1.0 4BvsMistral Large 4GPT-4.1 nanovsMistral Large 4Frequently asked questions
Nemotron Labs Dory 1.0 4B leads GPT-4.1 nano on 1 of the 1 benchmark they both report (GPQA Diamond). GPT-4.1 nano shipped 550 days before Nemotron Labs Dory 1.0 4B, so benchmark comparisons should account for the intervening progress.
Context windows are 256k (Nemotron Labs Dory 1.0 4B) vs 1M (GPT-4.1 nano). Nemotron Labs Dory 1.0 4B is open weight, while GPT-4.1 nano is closed.
On GPQA Diamond, Nemotron Labs Dory 1.0 4B leads at 72.41% vs GPT-4.1 nano at 50.3%.