DeepSeek V4.1 Flash vs GPT-6 Astra

At a glance

• Capability index

  • DeepSeek V4.1 Flash155.6
  • GPT-6 Astra166.4

  GPT-6 Astra is 10.8 points higher.

• Price per 1M tokens

  • DeepSeek V4.1 Flash$0.464
  • GPT-6 Astra$20.00

  DeepSeek V4.1 Flash is 43 times cheaper.

• Context window

  • DeepSeek V4.1 Flash1M tokens
  • GPT-6 Astra1.1M tokens

  GPT-6 Astra takes in 0% more.

• Released

  • DeepSeek V4.1 FlashSep 10, 2026
  • GPT-6 AstraSep 4, 2026

  DeepSeek V4.1 Flash is 6 days newer.

Where each scores higher

On the 21 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.

1 more test has results for only one of them; the table below lists every result.

Everything side by side

Comparison of DeepSeek V4.1 Flash vs GPT-6 Astra

Measure	DeepSeek V4.1 FlashDeepSeek	GPT-6 AstraOpenAI
Model
Maker	DeepSeek	OpenAI
Sold here by	engy	OpenAI
API model ID	deepseek-v4.1-flash	gpt-6-astra
Released	Sep 10, 2026	Sep 4, 2026
Knowledge cutoff	May 2025	Apr 30, 2026
Weights	Open: downloadable	Closed: hosted access only
Developer’s country	China	United States
Hosts selling it	80	35including OpenAI directly
Training compute	4.3 × 10²⁴ FLOPEpoch AI estimate	9.4 × 10²⁶ FLOPEpoch AI estimate
Test results
Capability index	155.6#31 of 277	166.4 (best of these models)#2 of 277
LiveBench overall	81.1%#9 of 70	82.2% (best of these models)#4 of 70
Arena text	1475#38 of 415	1477 (best of these models)#30 of 415
LiveBench coding	80.0%#23 of 70	80.4% (best of these models)#21 of 70
Arena WebDev	1619#24 of 142	1780 (best of these models)#2 of 142
Arena coding	1526#27 of 410	1543 (best of these models)#9 of 410
LiveBench mathematics	93.3%#23 of 70	96.8% (best of these models)#4 of 70
Mock AIME 2024–2025	98.3%#24 of 303	100.0% (best of these models)#1 of 303
FrontierMath Tiers 1–3	67.4%#30 of 117	93.7% (best of these models)#1 of 117
FrontierMath Tier 4	26.8%#41 of 73	97.6% (best of these models)#2 of 73
Arena math	1495 (best of these models)#22 of 400	1482#34 of 400
SimpleQA Verified	42.9%#43 of 90	75.6% (best of these models)#2 of 90
LiveBench agentic coding	77.3% (best of these models)#1 of 70	57.3%#22 of 70
DeepSWE	Not tested	74.1%#1 of 69
LiveBench language	81.2%#29 of 70	89.4% (best of these models)#4 of 70
LiveBench instructions	70.0%#32 of 70	75.6% (best of these models)#8 of 70
Arena creative writing	1439#61 of 413	1450 (best of these models)#39 of 413
Arena instruction following	1477 (best of these models)#24 of 415	1472#35 of 415
LiveBench data analysis	79.3%#18 of 70	83.0% (best of these models)#1 of 70
LiveBench reasoning	86.7%#32 of 70	92.7% (best of these models)#1 of 70
GPQA Diamond	89.8%#50 of 323	95.8% (best of these models)#2 of 323
Chess puzzles	48.0%#20 of 243	74.0% (best of these models)#1 of 243
Arena hard prompts	1499#36 of 415	1502 (best of these models)#30 of 415
Price per million tokens
Input	$0.04 (best of these models)	$10.00
Output	$0.08 (best of these models)	$50.00
Blended, 3 input to 1 output	$0.05 (best of these models)	$20.00
Cached input	$0.008 (best of these models)	$1.00
Writing to the cache	Not reported	$12.50
Long requests	No separate price reported	$20.00 in, $75.00 outabove 272K tokens
The maker’s own price	Not sold directly	This listing
Limits
Context window	327.7K tokens	1.1M tokens (best of these models)
Max input	1M tokens (best of these models)	922K tokens
Max output	65.5K tokens	128K tokens (best of these models)
Features
Reads	images and text	images, PDFs and text
Produces	text	text
Reasoning	Yeseffort none, low, high, max; can be switched off; thinking budget can be set	Yeseffort low, medium, high, xhigh, max
Tool calling	Yes	Yes
Structured output	Yes, such as JSON	Yes, such as JSON
File attachments	Yes	Yes
Temperature setting	Supported	Fixed by the provider

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.