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. | 49B | — |
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. | 128k | 400k |
| API pricingUSD per 1M tokens · lower wins | ||
Input priceWhat you pay for everything you send the model — your question, plus any documents or earlier conversation you include with it. | — | $0.05 |
Output priceWhat you pay for the text the model writes back. It is normally the dearer half: producing an answer costs more than reading one. | — | $0.40 |
Cached input priceA reduced rate for text you send over and over. If every request starts with the same instructions or the same document, the provider keeps a copy ready and charges less to read it again. | — | $0.005 |
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. | — | $0.025OpenAI |
Cheapest outputLowest output rate across every provider on OpenRouter, the lab included. May come from a different provider than the cheapest input. | — | $0.20OpenAI |
| — | $30.00OpenAI list price | |
| Benchmarks | ||
AIME 2025Reasoning — single-answer accuracy on the 2025 invitational mathematics exam without tools | 82.71% | 85.2% |
GPQA DiamondScience — Graduate-level science questions in biology, physics, and chemistry — hard enough that subject-matter PhDs score around 65%. Higher is better. | 71.97% | 71.2% |
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. | — | 54.7% |
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. | 73.58% | — |
Aider Polyglot (diff)Coding — editing code in six languages using the diff edit format | — | 48.4% |
τ²-bench · retailAgentic tasks — task success in the retail domain of the second-generation customer-service evaluation | — | 62.3% |
τ²-bench · airlineAgentic tasks — task success in the airline domain of the second-generation customer-service evaluation | — | 41% |
τ²-bench · telecomAgentic tasks — task success in the telecom domain of the second-generation customer-service evaluation | — | 35.5% |
Humanity's Last Exam · no toolsMultidisciplinary reasoning — Humanity's Last Exam — extremely hard expert questions across many subjects, written so you can't just look up the answer. “No tools” means the AI answers on its own. Higher is better. | — | 8.7% |
Humanity's Last Exam · text only, no toolsMultidisciplinary reasoning — accuracy on the 2,158 text-only questions without tools | 7.64% | — |
MATH-500Reasoning — accuracy on the 500-problem MATH subset | 97.4% | — |
AIME 2024Reasoning — single-answer accuracy on the 2024 invitational mathematics exam without tools | 87.5% | — |
MMLU-ProKnowledge — accuracy on the ten-option MMLU-Pro benchmark | 79.53% | — |
CharXiv Reasoning · with toolsChart reasoning — The same chart-and-figure reasoning test, run with the AI allowed to use tools — writing code to inspect the image, for instance — rather than reading the chart unaided. Scores run higher than the unaided version, so read the two as separate tests. Higher is better. | — | 62.7% |
MMMU-Pro · standard and vision meanMultimodal reasoning — average across the standard and vision sets | — | 62.6% |
MMMUMultimodal — Tests the AI on understanding images and text together across many college subjects. Higher is better. | — | 75.6% |
GraphWalks BFS (<128K)Long context — breadth-first traversal accuracy for inputs shorter than 128K tokens | — | 64% |
GraphWalks parents (<128K)Long context — parent-node retrieval accuracy for inputs shorter than 128K tokens | — | 43.8% |
OpenAI MRCR (2 needles) · 128KLong context — retrieval of two matching turns in the original MRCR evaluation | — | 43.2% |
OpenAI MRCR (2 needles) · 256KLong context — retrieval of two matching turns in the original MRCR evaluation | — | 34.9% |
| Overview | ||
| Company | NVIDIA | OpenAI |
| Release date | Jul 25 2025 | Aug 7 2025 |
| Access | Open Weight | Closed |
| Model details | View model | View model |
Other comparisons
Llama Nemotron Super 49B v1.5vsClaude Fable 5.2GPT-5 nanovsClaude Fable 5.2Llama Nemotron Super 49B v1.5vsGemini 4 NeonGPT-5 nanovsGemini 4 NeonLlama Nemotron Super 49B v1.5vsMuse EmberGPT-5 nanovsMuse EmberLlama Nemotron Super 49B v1.5vsGrok 4.7GPT-5 nanovsGrok 4.7Llama Nemotron Super 49B v1.5vsDeepSeek-V4.1-Flash-1017GPT-5 nanovsDeepSeek-V4.1-Flash-1017Llama Nemotron Super 49B v1.5vsMistral Large 4GPT-5 nanovsMistral Large 4Frequently asked questions
Llama Nemotron Super 49B v1.5 and GPT-5 nano are evenly matched across the 2 benchmarks they both report (AIME 2025, GPQA Diamond). Only GPT-5 nano has a verified first-party price for an available API: $0.05 per million input tokens and $0.40 per million output tokens. No currently available pay-as-you-go API rate is tracked for Llama Nemotron Super 49B v1.5. Llama Nemotron Super 49B v1.5 shipped 13 days before GPT-5 nano, so benchmark comparisons should account for the intervening progress.
Context windows are 128k (Llama Nemotron Super 49B v1.5) vs 400k (GPT-5 nano). Llama Nemotron Super 49B v1.5 is open weight, while GPT-5 nano is closed.
On AIME 2025, GPT-5 nano leads at 85.2% vs Llama Nemotron Super 49B v1.5 at 82.71%. On GPQA Diamond, Llama Nemotron Super 49B v1.5 leads at 71.97% vs GPT-5 nano at 71.2%.