DeepSeek V4 Pro 0813 vs Gemini 3.7 Flash

At a glance

• Capability index

  • DeepSeek V4 Pro 0813155.3
  • Gemini 3.7 Flash157.3

  Gemini 3.7 Flash is 2.0 points higher.

• Price per 1M tokens

  • DeepSeek V4 Pro 0813$1.98
  • Gemini 3.7 Flash$1.50

  Gemini 3.7 Flash is 24% cheaper.

• Context window

  • DeepSeek V4 Pro 08131M tokens
  • Gemini 3.7 Flash1M tokens

  They take in about the same.

• Released

  • DeepSeek V4 Pro 0813Aug 12, 2026
  • Gemini 3.7 FlashAug 13, 2026

  Gemini 3.7 Flash is 1 day newer.

Where each scores higher

On the 14 tests both have taken, each on its own scale. A gap of a point or two can sit within a test’s margin of error.

Level on Chess puzzles.

8 more tests have results for only one of them; the table below lists every result.

Everything side by side

Comparison of DeepSeek V4 Pro 0813 vs Gemini 3.7 Flash

Measure	DeepSeek V4 Pro 0813DeepSeek	Gemini 3.7 FlashGoogle
Model
Maker	DeepSeek	Google
Sold here by	CrofAI	Google
API model ID	deepseek-v4-pro-0813	gemini-3.7-flash
Released	Aug 12, 2026	Aug 13, 2026
Knowledge cutoff	Not reported	Mar 2026
Weights	Open: downloadable	Closed: hosted access only
Developer’s country	China	United States
Hosts selling it	42	29including Google directly
Training compute	9.7 × 10²⁴ FLOPEpoch AI estimate	Not published
Test results
Capability index	155.3#29 of 274	157.3 (best of these models)#16 of 274
LiveBench overall	77.4%#23 of 66	78.8% (best of these models)#16 of 66
Arena text	Not tested	1488#17 of 413
LiveBench coding	77.2%#38 of 66	78.9% (best of these models)#29 of 66
Arena WebDev	Not tested	1592#27 of 138
Arena coding	Not tested	1518#41 of 408
LiveBench mathematics	95.1% (best of these models)#16 of 66	93.5%#20 of 66
Mock AIME 2024–2025	98.6% (best of these models)#20 of 297	97.2%#31 of 297
FrontierMath Tiers 1–3	64.6%#32 of 114	71.6% (best of these models)#23 of 114
FrontierMath Tier 4	26.8%#39 of 70	36.6% (best of these models)#30 of 70
Arena math	Not tested	1503#12 of 396
SimpleQA Verified	52.9%#21 of 86	69.2% (best of these models)#9 of 86
LiveBench agentic coding	55.0%#28 of 66	58.3% (best of these models)#19 of 66
DeepSWE	Not tested	65.5%#25 of 69
LiveBench language	82.1%#26 of 66	85.5% (best of these models)#13 of 66
LiveBench instructions	67.7%#36 of 66	79.9% (best of these models)#2 of 66
Arena creative writing	Not tested	1493#5 of 411
Arena instruction following	Not tested	1485#17 of 413
LiveBench data analysis	79.2% (best of these models)#18 of 66	68.0%#55 of 66
LiveBench reasoning	85.8%#34 of 66	87.8% (best of these models)#25 of 66
GPQA Diamond	91.7%#29 of 319	94.8% (best of these models)#5 of 319
Chess puzzles	47.0%#14 of 227	47.0%#14 of 227
Arena hard prompts	Not tested	1508#19 of 413
Price per million tokens
Input	$0.35 (best of these models)	$0.75
Output	$0.80 (best of these models)	$3.75
Blended, 3 input to 1 output	$0.462 (best of these models)	$1.50
Cached input	$0.01 (best of these models)	$0.075
Audio input	Not reported	$0.75
The maker’s own price	Not sold directly	This listing
Limits
Context window	1M tokens	1M tokens
Max input	1M tokens (best of these models)	936K tokens
Max output	131.1K tokens (best of these models)	65.5K tokens
Features
Reads	text	audio, images, PDFs, text and video
Produces	text	text
Reasoning	Yeseffort none, low, high, max; can be switched off; thinking budget can be set	Yeseffort low, medium, high
Tool calling	Yes	Yes
Structured output	Yes, such as JSON	Yes, such as JSON
File attachments	No	Yes
Temperature setting	Supported	Supported

Bold marks the better value in each row. Prices are each model’s own list price, or the middle price across hosts when the maker does not sell it, blended as three input tokens for every output token. Test results as published by LiveBench (Apache-2.0), Epoch AI (CC BY 4.0), Datacurve DeepSWE, compiled by Epoch AI (CC BY 4.0), Arena (CC BY 4.0); model details and prices from models.dev (MIT). Confirm pricing with the provider before use.