Skip to content
Model Pareto

DeepSeek V4.1 Flash 1017

DeepSeekOpen weightshuggingface.co ↗
Compare with nearest rivalsCompare with…
Input
$0.30/M tok
Output
$1.20/M tok
Blended
$0.525/M tok · 3:1
Speed
~74tok/sest.
Context
1Mtokens

Dated snapshot (1017) of V4.1 Flash; a separate checkpoint from deepseek-v4-1-flash (0910), described as a minor update with gains in reasoning, coding, agentic and visual benchmarks. Open weights (MIT). Peak-hour price; off-peak is 50% off.

Rankings

Self-hostingEstimated

How we estimate →

What it would cost to run these open weights yourself on rented GPUs, compared with the API price above.

Hardware
8× B200 180GB
FP8 weights
Throughput
~4,098 tok/s
many requests batched
Price
$2.27–20.3 /M tok
busy → light use
On your own machine
Mac Studio M5 Ultra 512GB (int4, ~2 tok/s single-stream)
single consumer GPU or Mac
Assumptions (5)
  • FP8 weights (763B params) + 40% KV-cache headroom ≈ 1068 GB
  • 8× B200 180GB at $5.98–$14.24/GPU-hour on-demand (2026-09-24)
  • ~4,098 output tok/s aggregate at batch 117 (bandwidth-bound)
  • Blended 3:1 input:output; prefill ~11,147 tok/s
  • Low = 75% utilization at the low GPU price; high = 20% at the high price

Benchmark scores

~ italic, dashed = no published score yet, estimated from related benchmarks.

Coding

SWE-bench VerifiedEstimated
~#2 of 20 (estimated)~88.2%
  1. #1Claude Opus 596.0%
  2. ~#2DeepSeek V4.1 Flash 1017Estimated88.2%
  3. #2Claude Sonnet 585.2%
No published score — estimated from related benchmarks

Hard reasoning

GPQA DiamondEstimated
~#2 of 191 (estimated)~95.5%
  1. #1GPT-6 Astra96.1%
  2. ~#2DeepSeek V4.1 Flash 1017Estimated95.5%
  3. #2Gemini 3.8 Flash95.3%
No published score — estimated from related benchmarks
AIME (latest)Estimated
~#2 of 33 (estimated)~98.9%
  1. #1GLM-5.3 NVFP499.5%
  2. ~#2DeepSeek V4.1 Flash 1017Estimated98.9%
  3. #2Beam 1.0 501B A23B97.7%
No published score — estimated from related benchmarks