Across 11 comparable benchmarks, DeepSeek-V3 has better reported results on 8 and worse on 1, with 2 tied.
Strong evidence11 of 11 shared results countedHow we compare
Price basis: Standard rates vs listed rates. The price section below shows the applicable periods.
| Fact | DeepSeek-V3 | Llama 4 Maverick |
|---|---|---|
| Context window | 128K tokens | 1M tokens |
| Input / 1M | $0.320 | — |
| Output / 1M | $0.890 | — |
| Blended (4:1) | $0.434 | — |
| Status | Available | Available |
| Released | December 26, 2024 | April 5, 2025 |
Blended is 4:1 input to output with 90% cache hits at listed rates — set your own mix.
Mean of 5 shared benchmarks grouped by capability on a 0–100 scale. How averages are built
DeepSeek-V3 ahead on 1 of 2 counted
Llama 4 Maverick ahead on 1 of 1 counted
No fixed-scale average — see the rows in Benchmark results
DeepSeek-V3 ahead on 2 of 2 counted
DeepSeek-V3 ahead on 1 of 1 counted
No fixed-scale average — see the rows in Benchmark results
Tied on 1 counted
DeepSeek-V3 ahead on 1 of 1 counted
No fixed-scale average — see the rows in Benchmark results
No shared comparable results for Computer & tool use, Research & search, Long context, Cybersecurity.
Published scores, with matching modes where available. Gaps show DeepSeek-V3 relative to Llama 4 Maverick. Reading the results
| Benchmark | DeepSeek-V3 | Llama 4 Maverick | Gap |
|---|---|---|---|
| Aider-PolyglotAgentic Development · Accuracy | 48.4Normal · no tools | 15.6Normal · no tools | +32.8 pts#25 vs #42 of 47 |
| Terminal Bench HardAgentic Development · Accuracy | 6.8Normal · tools | 6.8Normal · tools | =#123 vs #123 of 157 |
| SimpleBenchCommonsense · Score (AVG@5) | 18.9Normal · no tools | 27.7Normal · no tools | 0.68×#102 vs #88 of 106 |
| FrontierMathMathematics · Accuracy | 1.7Normal · no tools | 0.7Normal · no tools | +1.0 pts#38 vs #43 of 48 |
| MATHMathematics · Accuracy | 87.8Normal · no tools | 61.2Normal · no tools | +26.6 pts#7 vs #30 of 42 |
| SciCodeScientific Computing · Score | 35.8Normal · no tools | 31.7Normal · no tools | 1.13×#86 vs #91 of 98 |
| GDP.pdfDocuments & Charts · Task pass rate (%) | 0.8Normal · no tools | 0.8Normal · no tools | =#93 vs #93 of 96 |
| τ³-BankingService Workflows · Score | 4.7Normal · tools | 3.7Normal · tools | 1.27×#113 vs #116 of 120 |
| MMLUKnowledge Exams · Accuracy | 88.5Normal · no tools | 85.5Normal · no tools | +3.0 pts#15 vs #34 of 114 |
| MMLU-ProKnowledge Exams · Accuracy | 75.9Normal · no tools | 62.9Normal · no tools | +13.0 pts#80 vs #108 of 153 |
| ECICapability Indices · ECI score (capability index, higher is better) | 132Reported best | 132Reported best | 1.00×#114 vs #115 of 168 |
Agentic Development · Accuracy
Agentic Development · Accuracy
Commonsense · Score (AVG@5)
Mathematics · Accuracy
Mathematics · Accuracy
Scientific Computing · Score
Documents & Charts · Task pass rate (%)
Service Workflows · Score
A dash means no result is available.
| Benchmark | DeepSeek-V3 | Llama 4 Maverick |
|---|---|---|
| AIME 2024 | 39.0 | — |
| GPQA Diamond | 59.1 | — |
| HLE | — | 5.7 |
| IMO-ProofBench Advanced | 4.3 | — |
| MATH-500 | 87.8 | — |
| HumanEval | 89.0 | — |
| LiveCodeBench | 34.6 | — |
| MBPP | — | 77.6 |
| Pinch Bench | — | 46.1 |
| MMMU | — | 73.4 |
| MMMU-Pro | — | 62.1 |
| ARC-AGI-1 | — | 4.4 |
| ARC-AGI-2 | — | 0.0 |
| BBH | 92.3 | — |
| SimpleQA | 24.9 | — |
Time-of-day providers use peak-hour rates; fixed-price providers use standard rates. Off-peak discounts remain in the published details.
Drag the handle above the bar for cache hits, the one below it for the input : output mix · 5% steps
Blended price = effective input rate at 90% cache hits × 80% + output rate × 20%
A blended cost needs a published input and output price on both sides, and one of these two does not have them yet.
DeepSeek-V3: Standard pricing. DeepInfra official models API (deepseek-ai/DeepSeek-V3) verified 2026-11-13
Llama 4 Maverick: Listed text rates; pricing tier not confirmed.
| Price / 1M tokens | DeepSeek-V3 | Llama 4 Maverick | Ratio |
|---|---|---|---|
| Input | $0.32 | — | — |
| Output | $0.89 | — | — |
| Cached input | — | — | — |
| Cache write | — | — | — |
| Blended | $0.434 | — | — |
Estimates use the selected token mix. Cache writes, tool fees and long-context surcharges are excluded.
Reported cost per task, paired with the score from the same run. Settings and sources may differ; costs do not affect the verdict.
| Attribute | DeepSeek-V3 | Llama 4 Maverick |
|---|---|---|
| Availability | Available | Available |
| Provider | DeepSeek-AI | Facebook AI Research Lab |
| Model type | Chat model | Multimodal model |
| First release | December 26, 2024 | April 5, 2025 |
| Context window | 128K tokens | 1M tokens |
| Max output | — | 4K tokens |
| Total parameters | 681B | 400B |
| Architecture | Dense | Dense |
| Documented capabilities | Multilingual | Multilingual |
| Input modalities | Text | Text, Image, Audio, Video |
| Output modalities | Text | Text |
| Weights | Open weights | Open weights |
| Licence | DEEPSEEK LICENSE AGREEMENT | Llama4 License |
| Commercial use | Commercial use allowed | Commercial use allowed |
These are reported-score comparisons, not controlled head-to-head tests.
What people found comparing these models, or using either one. Comments also appear on the models' own pages.
Used DeepSeek-V3 or Llama 4 Maverick? Sign in with Google to share what worked, what did not, and which API or setup you tested.
No community comments on either model yet.
Alternatives with the most benchmark overlap for this matchup.