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. | 671B | — |
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 | 1M |
| 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. | — | $2.00 |
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. | — | $8.00 |
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. | — | $0.50 |
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. | $0.256DeepInfra | $2.00Azure |
Cheapest outputLowest output rate across every provider on OpenRouter, the lab included. May come from a different provider than the cheapest input. | $0.712DeepInfra | $8.00Azure |
| $103StreamLake | $800OpenAI list price | |
| Benchmarks | ||
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. | 42% | 54.6% |
GPQA DiamondScience — Graduate-level science questions in biology, physics, and chemistry — hard enough that subject-matter PhDs score around 65%. Higher is better. | 59.1% | 66.3% |
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. | 88.5% | 90.2% |
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. | — | 14% |
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. | 40.5% | — |
Aider Polyglot (diff)Coding — editing code in six languages using the diff edit format | — | 52.9% |
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. | — | 94.5% |
IFEvalAgentic tasks — following verifiable instructions in generated responses | — | 87.4% |
MATH-500Reasoning — accuracy on the 500-problem MATH subset | — | 82.1% |
DROPReasoning — three-shot token F1 on questions requiring reasoning over paragraphs | — | 79.4 |
AIME 2024Reasoning — single-answer accuracy on the 2024 invitational mathematics exam without tools | — | 48.1% |
SimpleQAKnowledge — accuracy on short, fact-seeking questions in the original SimpleQA set | — | 41.6% |
MMMLUKnowledge — average accuracy on multilingual translations of MMLU | — | 87.3% |
MGSMMultilingual math — solving grade-school word problems translated into multiple languages | — | 86.9% |
CharXiv ReasoningChart reasoning — Can the AI read and reason about complex charts and figures, not just text? Higher is better. | — | 56.7% |
MathVistaMultimodal reasoning — mathematical reasoning about visual information | — | 72.2% |
MMMUMultimodal — Tests the AI on understanding images and text together across many college subjects. Higher is better. | — | 74.8% |
GraphWalks BFS (<128K)Long context — breadth-first traversal accuracy for inputs shorter than 128K tokens | — | 61.7% |
GraphWalks parents (<128K)Long context — parent-node retrieval accuracy for inputs shorter than 128K tokens | — | 58% |
OpenAI MRCR (2 needles) · 128KLong context — retrieval of two matching turns in the original MRCR evaluation | — | 57.2% |
OpenAI MRCR (2 needles) · 1MLong context — retrieval of two matching turns in the original MRCR evaluation | — | 46.3% |
| Overview | ||
| Company | DeepSeek | OpenAI |
| Release date | Dec 26 2024 | Apr 14 2025 |
| Access | Open Weight | Closed |
| Model details | View model | View model |
Other comparisons
DeepSeek V3vsClaude Fable 5.2GPT-4.1vsClaude Fable 5.2DeepSeek V3vsGemini 4 HydrogenGPT-4.1vsGemini 4 HydrogenDeepSeek V3vsMuse EmberGPT-4.1vsMuse EmberDeepSeek V3vsGrok 4.7GPT-4.1vsGrok 4.7DeepSeek V3vsMistral Large 4GPT-4.1vsMistral Large 4DeepSeek V3vsKimi K3.1 ProGPT-4.1vsKimi K3.1 ProFrequently asked questions
GPT-4.1 leads DeepSeek V3 on 3 of the 3 benchmarks they both report (SWE-Bench Verified, GPQA Diamond, MMLU). Only GPT-4.1 has a verified first-party price for an available API: $2.00 per million input tokens and $8.00 per million output tokens. No currently available pay-as-you-go API rate is tracked for DeepSeek V3. DeepSeek V3 shipped 109 days before GPT-4.1, so benchmark comparisons should account for the intervening progress.
Context windows are 128k (DeepSeek V3) vs 1M (GPT-4.1). DeepSeek V3 is open weight, while GPT-4.1 is closed.
On SWE-Bench Verified, GPT-4.1 leads at 54.6% vs DeepSeek V3 at 42%. On GPQA Diamond, GPT-4.1 leads at 66.3% vs DeepSeek V3 at 59.1%. On MMLU, GPT-4.1 leads at 90.2% vs DeepSeek V3 at 88.5%.