vs
0:10
overall capability wins (0 ties across 10 scored). DeepSeek V4 Pro 0813 leads overall.
- Reasoning0:4
- Coding0:2
- Tool use0:1
- Composite indices0:2
- Pricing3:0
- Specs0:3
Verdict
DeepSeek V4 Pro 0813 leads coding (widest gap: +61.8 on TermBench); DeepSeek V4 Pro 0813 offers ~7.8x larger context (1M vs 128K); DeepSeek V4 Pro 0813 allows ~46.9x more max output (384K vs 8K).
- DeepSeek V4 Pro 0813 leads coding (widest gap: +61.8 on TermBench)
- DeepSeek V4 Pro 0813 offers ~7.8x larger context (1M vs 128K)
- DeepSeek V4 Pro 0813 allows ~46.9x more max output (384K vs 8K)
- DeepSeek V3 ships open weights (MIT)
- DeepSeek V3 is ~4.1x cheaper on a blended token basis than DeepSeek V4 Pro 0813
Cost for 1M in + 250K out
Illustrative chat workload at primary-provider list prices.
- DeepSeek V3
- $0.55
- DeepSeek V4 Pro 0813
- $2.31
DeepSeek V3 is about 4.2x cheaper on this workload.
Price per 1M tokens
Full comparison
Organization
DeepSeek
DeepSeek
Family
DeepSeek
DeepSeek
License
MIT
Proprietary
Open weights
Yes
No
Release
Dec 26, 2024
Aug 13, 2026
Knowledge cutoff
-
-
API / provider
DeepSeek
DeepSeek
Modalities
text → text
text → text
Specs
Context window
128K
1M
Max output
8K
384K
Parameters
671B (37B act.)
1.6T (49B act.)
Pricing
Input $/1M
$0.27
$1.32
Output $/1M
$1.10
$3.96
Blended $/1M (3∶1)
$0.48
$1.98
Speed
tok/s
80
-
TTFT (s)
0.4
-
Reasoning
MMLU-Pro
75.9
—
GPQA Diamond
55.7
92.8
Humanity's Last Exam
2.9
41.0
AIME 2025
26.0
—
MATH-500
90.2
—
Humanity's Last Exam (with tools)
—
—
AA-Omniscience Accuracy
25.5
49.1
AA-LCR v1.1
29.3
80.3
CritPt
—
18.0
MMMU-Pro
—
—
IFBench
34.8
—
Chartography
—
—
Chartography (With Tools)
—
—
Coding
SWE-bench Verified
—
—
SWE-bench Pro
—
—
SWE-bench Multilingual
—
—
LiveCodeBench
49.6
—
Terminal-Bench 2.1
16.9
78.7
Aider Polyglot
55.1
—
Terminal-Bench 3
—
—
BigCodeBench
50.0
—
SciCode
35.8
51.0
CursorBench
—
—
SWE-Rebench
23.3
—
NL2Repo-Bench
—
61.5
DeepSWE
—
62.7
WebDev Arena
—
—
Terminal-Bench 4.0
—
—
CursorBench 4.0
—
—
FrontierCode v1.1 (Main)
—
—
Arena
LMArena Elo
1,358
—
| Metric | DeepSeek V3 | DeepSeek V4 Pro 0813 | Delta |
|---|---|---|---|
| Identity | |||
| Organization | DeepSeek | DeepSeek | — |
| Family | DeepSeek | DeepSeek | — |
| License | MIT | Proprietary | — |
| Open weights | Yes | No | — |
| Release | Dec 26, 2024 | Aug 13, 2026 | — |
| Knowledge cutoff | - | - | — |
| API / provider | DeepSeek | DeepSeek | — |
| Modalities | text → text | text → text | — |
| Specs | |||
| Context window | 128K | 1M | B +872K |
| Max output | 8K | 384K | B +376K |
| Parameters | 671B (37B act.) | 1.6T (49B act.) | B +929 |
| Pricing | |||
| Input $/1M | $0.27 | $1.32 | A +$1.05 |
| Output $/1M | $1.10 | $3.96 | A +$2.86 |
| Blended $/1M (3∶1) | $0.48 | $1.98 | A +$1.50 |
| Speed | |||
| tok/s | 80 | - | — |
| TTFT (s) | 0.4 | - | — |
| Reasoning | |||
| MMLU-Pro | 75.9 | — | — |
| GPQA Diamond | 55.7 | 92.8 | B +37.1 pts |
| Humanity's Last Exam | 2.9 | 41.0 | B +38.1 pts |
| AIME 2025 | 26.0 | — | — |
| MATH-500 | 90.2 | — | — |
| Humanity's Last Exam (with tools) | — | — | — |
| AA-Omniscience Accuracy | 25.5 | 49.1 | B +23.6 pts |
| AA-LCR v1.1 | 29.3 | 80.3 | B +51.0 pts |
| CritPt | — | 18.0 | — |
| MMMU-Pro | — | — | — |
| IFBench | 34.8 | — | — |
| Chartography | — | — | — |
| Chartography (With Tools) | — | — | — |
| Coding | |||
| SWE-bench Verified | — | — | — |
| SWE-bench Pro | — | — | — |
| SWE-bench Multilingual | — | — | — |
| LiveCodeBench | 49.6 | — | — |
| Terminal-Bench 2.1 | 16.9 | 78.7 | B +61.8 pts |
| Aider Polyglot | 55.1 | — | — |
| Terminal-Bench 3 | — | — | — |
| BigCodeBench | 50.0 | — | — |
| SciCode | 35.8 | 51.0 | B +15.2 pts |
| CursorBench | — | — | — |
| SWE-Rebench | 23.3 | — | — |
| NL2Repo-Bench | — | 61.5 | — |
| DeepSWE | — | 62.7 | — |
| WebDev Arena | — | — | — |
| Terminal-Bench 4.0 | — | — | — |
| CursorBench 4.0 | — | — | — |
| FrontierCode v1.1 (Main) | — | — | — |
| Arena | |||
| LMArena Elo | 1,358 | — | — |
DeepSeek V3
- Context
- 128K / 8K out
- Parameters
- 671B (37B act.)
- Price
- $0.27 / $1.10
- Speed
- 80 tok/s · 0.4s TTFT
- Modalities
- text → text
- License
- MIT
Original DeepSeek V3 MoE - the release that reset open-weight price/performance expectations.
DeepSeek V4 Pro 0813
- Context
- 1M / 384K out
- Parameters
- 1.6T (49B act.)
- Price
- $1.32 / $3.96 peak · $0.66 / $1.98 off-peak
- Speed
- —
- Modalities
- text → text
- License
- Proprietary
GA release of DeepSeek-V4-Pro (Aug 13, 2026) with substantially enhanced agentic coding (Terminal-Bench 2.1 87.9, NL2Repo 61.5, DeepSWE 62.7), native Responses API support, and low/high/max thinking-effort control. HLE is 42.7 without tools and 60.0 with tools. The hosted 0813 checkpoint weights are not yet published. Peak/off-peak API pricing took effect Aug 16, 2026: $1.32 in / $3.96 out peak, $0.66 in / $1.98 out off-peak (per 1M tokens).
Benchmark charts
Winner bars are emphasized. Per-benchmark deltas sit above each chart.