vs
0:7
overall capability wins (0 ties across 7 scored). DeepSeek V4 Flash 0731 leads overall.
- Reasoning0:3
- Coding0:1
- Pricing3:0
- Specs0:3
Verdict
DeepSeek V4 Flash 0731 leads coding (widest gap: +12.7 on SciCode); DeepSeek V4 Flash 0731 offers ~7.8x larger context (1M vs 128K); DeepSeek V4 Flash 0731 allows ~46.9x more max output (384K vs 8K).
- DeepSeek V4 Flash 0731 leads coding (widest gap: +12.7 on SciCode)
- DeepSeek V4 Flash 0731 offers ~7.8x larger context (1M vs 128K)
- DeepSeek V4 Flash 0731 allows ~46.9x more max output (384K vs 8K)
- DeepSeek-R1-Distill-Qwen-32B is ~2.2x cheaper on a blended token basis than DeepSeek V4 Flash 0731
Cost for 1M in + 250K out
Illustrative chat workload at primary-provider list prices.
- DeepSeek-R1-Distill-Qwen-32B
- $0.38
- DeepSeek V4 Flash 0731
- $0.77
DeepSeek-R1-Distill-Qwen-32B is about 2.1x cheaper on this workload.
Price per 1M tokens
Full comparison
Organization
DeepSeek
DeepSeek
Family
DeepSeek
DeepSeek
License
MIT
MIT
Open weights
Yes
Yes
Release
Jan 20, 2025
Jul 31, 2026
Knowledge cutoff
-
-
API / provider
Together / Fireworks (ref.)
DeepSeek
Modalities
text → text
text → text
Specs
Context window
128K
1M
Max output
8K
384K
Parameters
32B
304B
Pricing
Input $/1M
$0.30
$0.44
Output $/1M
$0.30
$1.32
Blended $/1M (3∶1)
$0.30
$0.66
Speed
tok/s
70
-
TTFT (s)
0.35
-
Reasoning
MMLU-Pro
73.9
—
GPQA Diamond
61.5
90.8
Humanity's Last Exam
4.6
38.6
AIME 2025
63.0
—
MATH-500
94.3
—
Humanity's Last Exam (with tools)
—
—
AA-Omniscience Accuracy
—
40.4
AA-LCR v1.1
8.7
79.7
CritPt
—
16.6
MMMU-Pro
—
—
IFBench
22.9
—
Chartography
—
—
Chartography (With Tools)
—
—
Coding
SWE-bench Verified
—
—
SWE-bench Pro
—
—
SWE-bench Multilingual
—
—
LiveCodeBench
57.2
—
Terminal-Bench 2.1
—
78.7
Aider Polyglot
—
—
Terminal-Bench 3
—
—
BigCodeBench
43.9
—
SciCode
37.6
50.3
CursorBench
—
—
SWE-Rebench
—
—
NL2Repo-Bench
—
54.2
DeepSWE
—
54.4
WebDev Arena
—
—
Terminal-Bench 4.0
—
—
CursorBench 4.0
—
—
FrontierCode v1.1 (Main)
—
—
Arena
LMArena Elo
—
—
| Metric | DeepSeek-R1-Distill-Qwen-32B | DeepSeek V4 Flash 0731 | Delta |
|---|---|---|---|
| Identity | |||
| Organization | DeepSeek | DeepSeek | — |
| Family | DeepSeek | DeepSeek | — |
| License | MIT | MIT | — |
| Open weights | Yes | Yes | — |
| Release | Jan 20, 2025 | Jul 31, 2026 | — |
| Knowledge cutoff | - | - | — |
| API / provider | Together / Fireworks (ref.) | DeepSeek | — |
| Modalities | text → text | text → text | — |
| Specs | |||
| Context window | 128K | 1M | B +872K |
| Max output | 8K | 384K | B +376K |
| Parameters | 32B | 304B | B +272 |
| Pricing | |||
| Input $/1M | $0.30 | $0.44 | A +$0.14 |
| Output $/1M | $0.30 | $1.32 | A +$1.02 |
| Blended $/1M (3∶1) | $0.30 | $0.66 | A +$0.36 |
| Speed | |||
| tok/s | 70 | - | — |
| TTFT (s) | 0.35 | - | — |
| Reasoning | |||
| MMLU-Pro | 73.9 | — | — |
| GPQA Diamond | 61.5 | 90.8 | B +29.3 pts |
| Humanity's Last Exam | 4.6 | 38.6 | B +34.0 pts |
| AIME 2025 | 63.0 | — | — |
| MATH-500 | 94.3 | — | — |
| Humanity's Last Exam (with tools) | — | — | — |
| AA-Omniscience Accuracy | — | 40.4 | — |
| AA-LCR v1.1 | 8.7 | 79.7 | B +71.0 pts |
| CritPt | — | 16.6 | — |
| MMMU-Pro | — | — | — |
| IFBench | 22.9 | — | — |
| Chartography | — | — | — |
| Chartography (With Tools) | — | — | — |
| Coding | |||
| SWE-bench Verified | — | — | — |
| SWE-bench Pro | — | — | — |
| SWE-bench Multilingual | — | — | — |
| LiveCodeBench | 57.2 | — | — |
| Terminal-Bench 2.1 | — | 78.7 | — |
| Aider Polyglot | — | — | — |
| Terminal-Bench 3 | — | — | — |
| BigCodeBench | 43.9 | — | — |
| SciCode | 37.6 | 50.3 | B +12.7 pts |
| CursorBench | — | — | — |
| SWE-Rebench | — | — | — |
| NL2Repo-Bench | — | 54.2 | — |
| DeepSWE | — | 54.4 | — |
| WebDev Arena | — | — | — |
| Terminal-Bench 4.0 | — | — | — |
| CursorBench 4.0 | — | — | — |
| FrontierCode v1.1 (Main) | — | — | — |
| Arena | |||
| LMArena Elo | — | — | — |
DeepSeek-R1-Distill-Qwen-32B
- Context
- 128K / 8K out
- Parameters
- 32B
- Price
- $0.30 / $0.30
- Speed
- 70 tok/s · 0.35s TTFT
- Modalities
- text → text
- License
- MIT
Popular 32B R1 distill on Qwen — strong open STEM reasoning at mid-size GPU budgets.
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).
Benchmark charts
Winner bars are emphasized. Per-benchmark deltas sit above each chart.