// head_to_head

DeepSeek V4 Flash 0731 vs DeepSeek V4 Pro 0423

Published LiveBench scores across all seven categories, live list pricing, context windows, and the measured cost of a point of capability — for both models, side by side.

OpenRouter + LiveBenchAll comparisonsFull leaderboard

DeepSeek V4 Flash 0731 wins outright — it scores higher and costs less.

DeepSeek V4 Flash 0731 leads by 2.6 points overall while listing 6.3× cheaper per blended million tokens. There is no trade-off to reason about here; the only reason to pick the other is a constraint this table does not show, like an existing contract, a region, or a provider you are already on. The two cost measures disagree here, which is worth knowing: DeepSeek V4 Flash 0731 has the lower sticker price, but DeepSeek V4 Pro 0423 earns each point of capability for less — $0.0261 against $0.0356 — because per-token rates do not predict how many tokens a model actually spends on a task.

deepseek

DeepSeek V4 Flash 0731

Blended / 1M
$0.105
Context
1.3M
Released
Jul 31, 2026
Overall score
74.2
reasoningtool callingprompt caching

deepseek

DeepSeek V4 Pro 0423

Blended / 1M
$0.665
Context
1.0M
Released
Apr 24, 2026
Overall score
71.6
reasoningtool callingprompt caching

Specs and pricing

MetricDeepSeek V4 Flash 0731DeepSeek V4 Pro 0423
LiveBench overall

Mean of the seven LiveBench category scores, 0–100. Higher is better.

74.2win71.6
Cost per point

Measured benchmark spend divided by overall score — dollars per point of capability.

$0.0356$0.0261win
Blended price / 1M

3:1 input:output mix, the usual shape of production traffic.

$0.105win$0.665
Input price / 1M$0.080win$0.532
Output price / 1M$0.180win$1.06
Cached input / 1M

Price of an input token served from the prompt cache, where the provider publishes one.

$0.016win$0.044
Context window1.3M1.0M
Max output tokens384K384K

Benchmarks by category

An overall score averages away the thing you probably care about. These are the seven LiveBench categories scored separately — DeepSeek V4 Flash 0731 on top, DeepSeek V4 Pro 0423 below, both out of 100.

Agentic coding
46.8
42.6
Coding
75.0
70.0
Reasoning
86.6
82.7
Mathematics
86.8
90.7
Data analysis
79.3
74.5
Language
79.2
78.1
Instruction following
65.5
62.4

What each one costs to run

Per-token prices are hard to feel. These are monthly list costs for both models across five workload shapes, using each provider's published cached-input rate where there is one.

WorkloadDeepSeek V4 Flash 0731DeepSeek V4 Pro 0423
Support chatbot

1.2K in / 400 out × 200K requests

$28.99/mo$177.72/mo
RAG assistant

8K in / 600 out × 100K requests

$49.20/mo$294.42/mo
Coding agent

40K in / 4K out × 20K requests

$42.56/mo$237.67/mo
Document extraction

20K in / 1.5K out × 50K requests

$90.30/mo$587.52/mo
Bulk classification

500 in / 20 out × 5M requests

$186.00/mo$1192.77/mo
Run these two through the cost calculator

Which should you pick?

You are running this at volume

DeepSeek V4 Pro 0423

Lowest measured cost per point of capability at $0.0261 per point — the gap compounds with every request.

Quality matters more than the bill

DeepSeek V4 Flash 0731

Highest overall LiveBench score of the two at 74.2.

The workload is coding or agentic work

DeepSeek V4 Flash 0731

Leads on agentic coding — 46.8 against 42.6.

DeepSeek V4 Flash 0731 vs DeepSeek V4 Pro 0423 FAQ

Which is better, DeepSeek V4 Flash 0731 or DeepSeek V4 Pro 0423?

DeepSeek V4 Flash 0731 wins outright — it scores higher and costs less. DeepSeek V4 Flash 0731 leads by 2.6 points overall while listing 6.3× cheaper per blended million tokens. There is no trade-off to reason about here; the only reason to pick the other is a constraint this table does not show, like an existing contract, a region, or a provider you are already on. The two cost measures disagree here, which is worth knowing: DeepSeek V4 Flash 0731 has the lower sticker price, but DeepSeek V4 Pro 0423 earns each point of capability for less — $0.0261 against $0.0356 — because per-token rates do not predict how many tokens a model actually spends on a task.

Is DeepSeek V4 Flash 0731 cheaper than DeepSeek V4 Pro 0423?

DeepSeek V4 Flash 0731 is cheaper. On a 3:1 input:output blend, DeepSeek V4 Flash 0731 lists at $0.105 per million tokens and DeepSeek V4 Pro 0423 at $0.665 — DeepSeek V4 Flash 0731 is 6.3× cheaper. Input and output are priced separately — DeepSeek V4 Flash 0731 charges $0.080 in and $0.180 out, DeepSeek V4 Pro 0423 charges $0.532 and $1.06 — so the model that looks cheaper flips depending on how output-heavy your workload is.

DeepSeek V4 Flash 0731 vs DeepSeek V4 Pro 0423: which scores higher on benchmarks?

DeepSeek V4 Flash 0731 scores 74.2 and DeepSeek V4 Pro 0423 scores 71.6 overall on LiveBench, the mean of its seven categories. That is a 2.6-point lead for DeepSeek V4 Flash 0731. Category scores differ from the overall figure — a model can lead on reasoning and trail on coding, which the per-category table above breaks out.

Which gives better value for money, DeepSeek V4 Flash 0731 or DeepSeek V4 Pro 0423?

DeepSeek V4 Pro 0423. Cost per point divides the measured dollars LiveBench spent running the benchmark by the score it earned, so it captures something token pricing misses: a reasoning model can emit many times more tokens than its per-token rate suggests. DeepSeek V4 Flash 0731 works out at $0.0356 per point and DeepSeek V4 Pro 0423 at $0.0261.

Does DeepSeek V4 Flash 0731 or DeepSeek V4 Pro 0423 have a bigger context window?

They are effectively the same — 1.3M for DeepSeek V4 Flash 0731 and 1.0M for DeepSeek V4 Pro 0423.

Do DeepSeek V4 Flash 0731 and DeepSeek V4 Pro 0423 support prompt caching?

Both publish a cached-input rate: $0.016 per million for DeepSeek V4 Flash 0731 and $0.044 for DeepSeek V4 Pro 0423, against full input rates of $0.080 and $0.532. On a workload with a long stable prefix — a system prompt, a tool schema, a retrieved corpus — that changes the economics more than the headline price does.

Related comparisons

How these numbers are produced

  • Price — provider list price from OpenRouter, refreshed every 15 minutes. “Blended” is a 3:1 input:output mix.
  • ScoresLiveBench release 2026-06-25, using their own category map. Each model shows its strongest published run. A blank means “not evaluated”, never “bad”.
  • Cost per point — the measured dollars LiveBench spent on the run, divided by the score it earned.
  • “Win” — awarded only past a threshold: one full point on a benchmark score, 10% on a price, 25% on a context window. Anything tighter reports as a tie, because effort settings alone move a LiveBench score by more than that.

Published benchmarks rank models on someone else's tasks. Before committing, see LLM & agent evaluation for building an eval on your own.