vs
0:15
overall capability wins (3 ties across 18 scored). DeepSeek V4 Pro 0813 leads overall.
- Reasoning0:5
- Coding0:3
- Agents0:4
- Tool use0:2
- Composite indices0:2
- Pricing3:0
- Specs0:1
Verdict
DeepSeek V4 Pro 0813 leads coding (widest gap: +8.3 on DeepSWE); DeepSeek V4 Flash 0731 ships open weights (MIT); DeepSeek V4 Flash 0731 is ~3.0x cheaper on a blended token basis than DeepSeek V4 Pro 0813.
- DeepSeek V4 Pro 0813 leads coding (widest gap: +8.3 on DeepSWE)
- DeepSeek V4 Flash 0731 ships open weights (MIT)
- DeepSeek V4 Flash 0731 is ~3.0x 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 V4 Flash 0731
- $0.77
- DeepSeek V4 Pro 0813
- $2.31
DeepSeek V4 Flash 0731 is about 3.0x cheaper on this workload.
Price per 1M tokens
Full comparison
Organization
DeepSeek
DeepSeek
Family
DeepSeek
DeepSeek
License
MIT
Proprietary
Open weights
Yes
No
Release
Jul 31, 2026
Aug 13, 2026
Knowledge cutoff
-
-
API / provider
DeepSeek
DeepSeek
Modalities
text → text
text → text
Specs
Context window
1M
1M
Max output
384K
384K
Parameters
304B
1.6T (49B act.)
Pricing
Input $/1M
$0.44
$1.32
Output $/1M
$1.32
$3.96
Blended $/1M (3∶1)
$0.66
$1.98
Speed
tok/s
-
-
TTFT (s)
-
-
Reasoning
MMLU-Pro
—
—
GPQA Diamond
90.8
92.8
Humanity's Last Exam
38.6
41.0
AIME 2025
—
—
MATH-500
—
—
Humanity's Last Exam (with tools)
—
—
AA-Omniscience Accuracy
40.4
49.1
AA-LCR v1.1
79.7
80.3
CritPt
16.6
18.0
MMMU-Pro
—
—
IFBench
—
—
Chartography
—
—
Chartography (With Tools)
—
—
Coding
SWE-bench Verified
—
—
SWE-bench Pro
—
—
SWE-bench Multilingual
—
—
LiveCodeBench
—
—
Terminal-Bench 2.1
78.7
78.7
Aider Polyglot
—
—
Terminal-Bench 3
—
—
BigCodeBench
—
—
SciCode
50.3
51.0
CursorBench
—
—
SWE-Rebench
—
—
NL2Repo-Bench
54.2
61.5
DeepSWE
54.4
62.7
WebDev Arena
—
—
Terminal-Bench 4.0
—
—
CursorBench 4.0
—
—
FrontierCode v1.1 (Main)
—
—
Arena
LMArena Elo
—
—
| Metric | DeepSeek V4 Flash 0731 | DeepSeek V4 Pro 0813 | Delta |
|---|---|---|---|
| Identity | |||
| Organization | DeepSeek | DeepSeek | — |
| Family | DeepSeek | DeepSeek | — |
| License | MIT | Proprietary | — |
| Open weights | Yes | No | — |
| Release | Jul 31, 2026 | Aug 13, 2026 | — |
| Knowledge cutoff | - | - | — |
| API / provider | DeepSeek | DeepSeek | — |
| Modalities | text → text | text → text | — |
| Specs | |||
| Context window | 1M | 1M | tie |
| Max output | 384K | 384K | tie |
| Parameters | 304B | 1.6T (49B act.) | B +1K |
| Pricing | |||
| Input $/1M | $0.44 | $1.32 | A +$0.88 |
| Output $/1M | $1.32 | $3.96 | A +$2.64 |
| Blended $/1M (3∶1) | $0.66 | $1.98 | A +$1.32 |
| Speed | |||
| tok/s | - | - | — |
| TTFT (s) | - | - | — |
| Reasoning | |||
| MMLU-Pro | — | — | — |
| GPQA Diamond | 90.8 | 92.8 | B +2.0 pts |
| Humanity's Last Exam | 38.6 | 41.0 | B +2.4 pts |
| AIME 2025 | — | — | — |
| MATH-500 | — | — | — |
| Humanity's Last Exam (with tools) | — | — | — |
| AA-Omniscience Accuracy | 40.4 | 49.1 | B +8.7 pts |
| AA-LCR v1.1 | 79.7 | 80.3 | B +0.6 pts |
| CritPt | 16.6 | 18.0 | B +1.4 pts |
| MMMU-Pro | — | — | — |
| IFBench | — | — | — |
| Chartography | — | — | — |
| Chartography (With Tools) | — | — | — |
| Coding | |||
| SWE-bench Verified | — | — | — |
| SWE-bench Pro | — | — | — |
| SWE-bench Multilingual | — | — | — |
| LiveCodeBench | — | — | — |
| Terminal-Bench 2.1 | 78.7 | 78.7 | tie |
| Aider Polyglot | — | — | — |
| Terminal-Bench 3 | — | — | — |
| BigCodeBench | — | — | — |
| SciCode | 50.3 | 51.0 | B +0.7 pts |
| CursorBench | — | — | — |
| SWE-Rebench | — | — | — |
| NL2Repo-Bench | 54.2 | 61.5 | B +7.3 pts |
| DeepSWE | 54.4 | 62.7 | B +8.3 pts |
| WebDev Arena | — | — | — |
| Terminal-Bench 4.0 | — | — | — |
| CursorBench 4.0 | — | — | — |
| FrontierCode v1.1 (Main) | — | — | — |
| Arena | |||
| LMArena Elo | — | — | — |
DeepSeek V4 Flash 0731
- Context
- 1M / 384K out
- Parameters
- 304B
- Price
- $0.44 / $1.32 peak · $0.22 / $0.66 off-peak
- Speed
- —
- Modalities
- text → text
- License
- MIT
Official DeepSeek-V4-Flash release (Jul 31, 2026 public beta) superseding the preview - same base architecture plus a DSpark speculative-decoding module (304B total), MIT weights on Hugging Face, and agent benchmarks well above V4-Pro-Preview (Terminal-Bench 2.1 82.7, NL2Repo 54.2). Peak/off-peak API pricing took effect Aug 16, 2026: $0.44 in / $1.32 out peak, $0.22 in / $0.66 out off-peak (per 1M tokens).
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.