Superbash
SearchUser Hub

DeepSeek

Tier B · Specialist or second-choice models.

DeepSeek V4 Flash

Use DeepSeek V4 Flash for cheap planning, scaffolding, and rough implementation that will receive a full review.

API pricing

USD per million tokens · DeepSeek API peak rate
Input
$0.44
Output
$1.32
Cached input
$0.014
Context window
1M
Pricing details and specifications

Try the visual runs

Open the generated scene to test it, or compare that prompt with another model.

11 runs

Mechanical Watch Simulator

Interactive watch movement testing mechanical legibility, accurate relative motion, and real-time 3D controls.

Open run →Compare this model →

Stormwind Trebuchet Simulator

Counterweight siege simulation testing coupled mechanics, trajectory prediction, projectile cameras, and interactive tuning.

Open run →Compare this model →

Editorial verdict

Where this model fits

Back to all benchmarks →

DeepSeek V4 Flash is the budget choice in the August roster. The team spent about $1 on the shared visual set, while some other model runs cost roughly $25 to $40, but several outputs still showed visible defects.

Best for

  • Cheap daily implementation
  • Plan-mode brainstorming
  • Scaffolds and bulk rough drafts
  • Workloads that may need instant credit top-ups

Watch out

  • Visible visual bugs appeared in benchmark output
  • The lightweight model remains below Pro on capability
  • Requires review before production

Why it is ranked here

  1. The recorded spend was about $1 for the shared runs. Prompts, retries, and provider pricing can change that figure.
  2. Visible defects remained, including broken spatial composition in the Helm’s Deep run.
  3. The B-tier placement rewards cost and refillable capacity, not flagship-quality output.

Evidence

2026-09-07

Superbash editorial model ranking

Takeaway: DeepSeek V4 Flash is currently placed in Tier B.

The September 2026 editorial roster places DeepSeek V4 Flash at rank 11.

Open source →
2026-06-17

NEW AI Model Tier List for Vibe Coding!

Takeaway: Use DeepSeek V4 Flash for low-cost planning and scaffolding.

The historical source emphasized hand-holding and review; the current editorial roster raises it to B tier.

Open source →

Related guides

How DeepSeek V4 Flash compares

Three relevant peers, side by side.

  • DeepSeek V4 Flash This model
  • DeepSeek V4 Pro Max
  • Gemini 3.1 Pro High
  • Claude Opus 4.6 Max

6 benchmarks · 4 models

MMLU-Pro

Knowledge · Higher is better

Source
DeepSeek V4 Flash86.2%
DeepSeek V4 Pro Max87.5%
Gemini 3.1 Pro High91.0%
Claude Opus 4.6 Max89.1%

GPQA Diamond

STEM reasoning · Higher is better

Source
DeepSeek V4 Flash88.1%
DeepSeek V4 Pro Max90.1%
Gemini 3.1 Pro High94.3%
Claude Opus 4.6 Max91.3%

LiveCodeBench

Code reasoning · Higher is better

Source
DeepSeek V4 Flash91.6%
DeepSeek V4 Pro Max93.5%
Gemini 3.1 Pro High91.7%
Claude Opus 4.6 Max88.8%

HMMT February 2026

Math · Higher is better

Source
DeepSeek V4 Flash94.8%
DeepSeek V4 Pro Max95.2%
Gemini 3.1 Pro HighNot reported
Claude Opus 4.6 Max96.2%

IMOAnswerBench

Math reasoning · Higher is better

Source
DeepSeek V4 Flash88.4%
DeepSeek V4 Pro Max89.8%
Gemini 3.1 Pro HighNot reported
Claude Opus 4.6 MaxNot reported

MRCR 1M

Long context · Higher is better

Source
DeepSeek V4 Flash78.7%
DeepSeek V4 Pro Max83.5%
Gemini 3.1 Pro High76.3%
Claude Opus 4.6 Max92.9%

Official benchmark profile

Published scores outside our visual tests

DeepSeek sourceApril 2026Source report →
Knowledge

MMLU-Pro

86.2%
STEM reasoning

GPQA Diamond

88.1%
Code reasoning

LiveCodeBench

91.6%
Full official benchmark table12 scores and peer comparisons
BenchmarkAreaScoreComparison
MMLU-ProKnowledge86.2%
DeepSeek V4 Flash86.2%
DeepSeek V4 Pro Max87.5%
GPT-5.4 xHigh87.5%
Claude Opus 4.6 Max89.1%
GPQA DiamondSTEM reasoning88.1%
DeepSeek V4 Flash88.1%
DeepSeek V4 Pro Max90.1%
Claude Opus 4.6 Max91.3%
GPT-5.4 xHigh93.0%
LiveCodeBenchCode reasoning91.6%
DeepSeek V4 Flash91.6%
Gemini 3.1 Pro High91.7%
DeepSeek V4 Pro Max93.5%
Kimi K2.6 Thinking89.6%
HMMT February 2026Math94.8%
DeepSeek V4 Flash94.8%
DeepSeek V4 Pro Max95.2%
Claude Opus 4.6 Max96.2%
GPT-5.4 xHigh97.7%
IMOAnswerBenchMath reasoning88.4%
DeepSeek V4 Flash88.4%
DeepSeek V4 Pro Max89.8%
Kimi K2.6 Thinking86.0%
GPT-5.4 xHigh91.4%
MRCR 1MLong context78.7%
DeepSeek V4 Flash78.7%
Gemini 3.1 Pro High76.3%
DeepSeek V4 Pro Max83.5%
Claude Opus 4.6 Max92.9%
Terminal-Bench 2.0Agentic coding56.9%
DeepSeek V4 Flash56.9%
Kimi K2.6 Thinking66.7%
DeepSeek V4 Pro Max67.9%
Gemini 3.1 Pro High68.5%
SWE-bench VerifiedAgentic coding79.0%
DeepSeek V4 Flash79.0%
DeepSeek V4 Pro Max80.6%
Gemini 3.1 Pro High80.6%
Claude Opus 4.6 Max80.8%
SWE-bench ProAgentic coding52.6%
DeepSeek V4 Flash52.6%
DeepSeek V4 Pro Max55.4%
GPT-5.4 xHigh57.7%
GLM 5.1 Thinking58.4%
BrowseCompWeb research73.2%
DeepSeek V4 Flash73.2%
DeepSeek V4 Pro Max83.4%
Claude Opus 4.6 Max83.7%
Gemini 3.1 Pro High85.9%
MCP AtlasTool use69.0%
DeepSeek V4 Flash69.0%
Gemini 3.1 Pro High69.2%
GLM 5.1 Thinking71.8%
DeepSeek V4 Pro Max73.6%
ToolathlonTool use47.8%
DeepSeek V4 Flash47.8%
Kimi K2.6 Thinking50.0%
DeepSeek V4 Pro Max51.8%
GPT-5.4 xHigh54.6%
Sources, pricing details and methodology

Verified model facts

Model specifications

Pricing source · checked 2026-09-08 ↗
Status
Current
API model ID
deepseek-v4-flash
Context
1M
Max output
384K
License
Not publicly verified
Access route
Not publicly verified
Modalities
Not publicly verified
Hardware
Not publicly verified
DeepSeek API peak rate · USD / 1M tokens
$0.44 input · $0.014 cached input · $1.32 output

Peak hours: Monday–Friday, 01:00–04:00 and 06:00–10:00 UTC. All other hours use half-price off-peak rates: $0.22 input, $0.007 cached input, and $0.66 output per 1M tokens.

Benchmark sources

DeepSeek V4 Flash activates 13B of 284B parameters with a 1M-token context. These are the official V4-Flash Max results, which use the largest published reasoning budget and differ from the non-thinking and high settings.

DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence

Evaluation settings

  • Exact match at V4-Flash Max reasoning effort.
  • Pass@1 at V4-Flash Max reasoning effort.
  • Mean matching rate at the 1M-token setting.
  • Accuracy in the official DeepSeek comparison table.
  • Resolved rate in the official DeepSeek comparison table.
  • Pass@1 in the official DeepSeek comparison table.