vs
5:14
overall capability wins (2 ties across 21 scored). DeepSeek V4 Pro leads overall.
- Reasoning2:4
- Coding2:4
- Arena0:1
- Agents0:3
- Tool use1:1
- Composite indices0:2
- Pricing3:0
- Speed2:0
- Specs0:1
Verdict
DeepSeek V4 Pro leads coding (widest gap: +5.5 on DeepSWE); DeepSeek V4 Pro ranks higher on LMArena (+22 Elo); DeepSeek V4 Flash is ~3.1x cheaper on a blended token basis than DeepSeek V4 Pro.
- DeepSeek V4 Pro leads coding (widest gap: +5.5 on DeepSWE)
- DeepSeek V4 Pro ranks higher on LMArena (+22 Elo)
- DeepSeek V4 Flash is ~3.1x cheaper on a blended token basis than DeepSeek V4 Pro
Cost for 1M in + 250K out
Illustrative chat workload at primary-provider list prices.
- DeepSeek V4 Flash
- $0.21
- DeepSeek V4 Pro
- $0.65
DeepSeek V4 Flash is about 3.1x cheaper on this workload.
Price per 1M tokens
Full comparison
Organization
DeepSeek
DeepSeek
Family
DeepSeek
DeepSeek
License
MIT
MIT
Open weights
Yes
Yes
Release
Apr 24, 2026
Apr 24, 2026
Knowledge cutoff
-
-
API / provider
DeepSeek
DeepSeek
Modalities
text → text
text → text
Specs
Context window
1M
1M
Max output
384K
384K
Parameters
284B (13B act.)
1.6T (49B act.)
Pricing
Input $/1M
$0.14
$0.43
Output $/1M
$0.28
$0.87
Blended $/1M (3∶1)
$0.18
$0.54
Speed
tok/s
120
85
TTFT (s)
0.25
0.4
Reasoning
MMLU-Pro
—
—
GPQA Diamond
89.4
88.8
Humanity's Last Exam
34.8
37.5
AIME 2025
—
—
MATH-500
—
—
Humanity's Last Exam (with tools)
—
—
AA-Omniscience Accuracy
36.8
43.0
AA-LCR v1.1
74.3
74.7
CritPt
7.1
12.9
MMMU-Pro
—
—
IFBench
79.2
76.5
Chartography
—
—
Chartography (With Tools)
—
—
Coding
SWE-bench Verified
—
—
SWE-bench Pro
—
—
SWE-bench Multilingual
—
—
LiveCodeBench
—
—
Terminal-Bench 2.1
61.8
64.0
Aider Polyglot
—
—
Terminal-Bench 3
—
—
BigCodeBench
—
—
SciCode
45.3
50.8
CursorBench
—
—
SWE-Rebench
38.4
41.4
NL2Repo-Bench
39.4
38.5
DeepSWE
7.3
12.8
WebDev Arena
1,580
1,446
Terminal-Bench 4.0
—
—
CursorBench 4.0
—
—
FrontierCode v1.1 (Main)
—
—
Arena
LMArena Elo
1,436
1,457
| Metric | DeepSeek V4 Flash | DeepSeek V4 Pro | Delta |
|---|---|---|---|
| Identity | |||
| Organization | DeepSeek | DeepSeek | — |
| Family | DeepSeek | DeepSeek | — |
| License | MIT | MIT | — |
| Open weights | Yes | Yes | — |
| Release | Apr 24, 2026 | Apr 24, 2026 | — |
| Knowledge cutoff | - | - | — |
| API / provider | DeepSeek | DeepSeek | — |
| Modalities | text → text | text → text | — |
| Specs | |||
| Context window | 1M | 1M | tie |
| Max output | 384K | 384K | tie |
| Parameters | 284B (13B act.) | 1.6T (49B act.) | B +1K |
| Pricing | |||
| Input $/1M | $0.14 | $0.43 | A +$0.29 |
| Output $/1M | $0.28 | $0.87 | A +$0.59 |
| Blended $/1M (3∶1) | $0.18 | $0.54 | A +$0.37 |
| Speed | |||
| tok/s | 120 | 85 | A +35 |
| TTFT (s) | 0.25 | 0.4 | A +0.15 |
| Reasoning | |||
| MMLU-Pro | — | — | — |
| GPQA Diamond | 89.4 | 88.8 | A +0.6 pts |
| Humanity's Last Exam | 34.8 | 37.5 | B +2.7 pts |
| AIME 2025 | — | — | — |
| MATH-500 | — | — | — |
| Humanity's Last Exam (with tools) | — | — | — |
| AA-Omniscience Accuracy | 36.8 | 43.0 | B +6.2 pts |
| AA-LCR v1.1 | 74.3 | 74.7 | B +0.4 pts |
| CritPt | 7.1 | 12.9 | B +5.8 pts |
| MMMU-Pro | — | — | — |
| IFBench | 79.2 | 76.5 | A +2.7 pts |
| Chartography | — | — | — |
| Chartography (With Tools) | — | — | — |
| Coding | |||
| SWE-bench Verified | — | — | — |
| SWE-bench Pro | — | — | — |
| SWE-bench Multilingual | — | — | — |
| LiveCodeBench | — | — | — |
| Terminal-Bench 2.1 | 61.8 | 64.0 | B +2.2 pts |
| Aider Polyglot | — | — | — |
| Terminal-Bench 3 | — | — | — |
| BigCodeBench | — | — | — |
| SciCode | 45.3 | 50.8 | B +5.5 pts |
| CursorBench | — | — | — |
| SWE-Rebench | 38.4 | 41.4 | B +3.0 pts |
| NL2Repo-Bench | 39.4 | 38.5 | A +0.9 pts |
| DeepSWE | 7.3 | 12.8 | B +5.5 pts |
| WebDev Arena | 1,580 | 1,446 | A +135 Elo |
| Terminal-Bench 4.0 | — | — | — |
| CursorBench 4.0 | — | — | — |
| FrontierCode v1.1 (Main) | — | — | — |
| Arena | |||
| LMArena Elo | 1,436 | 1,457 | B +22 Elo |
DeepSeek V4 Flash
- Context
- 1M / 384K out
- Parameters
- 284B (13B act.)
- Price
- $0.14 / $0.28
- Speed
- 120 tok/s · 0.25s TTFT
- Modalities
- text → text
- License
- MIT
Budget DeepSeek V4 API tier (Apr 2026 preview). Superseded by DeepSeek-V4-Flash-0731 (public beta), which re-post-trains the same architecture and reports far stronger agent benchmarks.
DeepSeek V4 Pro
- Context
- 1M / 384K out
- Parameters
- 1.6T (49B act.)
- Price
- $0.43 / $0.87
- Speed
- 85 tok/s · 0.4s TTFT
- Modalities
- text → text
- License
- MIT
Open-weight frontier MoE - exceptional quality-per-dollar with MIT license and 1M context. Scores shown are the original preview; the GA 0813 release improves agentic coding substantially.
Benchmark charts
Winner bars are emphasized. Per-benchmark deltas sit above each chart.