Open Source AI

DeepSeek V4 Flash Tested: Tiny API Cost, Strong Coding, Huge Local Footprint

Direct Answer

DeepSeek V4 Flash 0731 is one of the most interesting low-cost coding models available, but “small” needs an asterisk. The official model has 284 billion total parameters, activates about 13 billion per token, supports a one-million-token context window, and publishes its weights under MIT. In Bijan Bowen's eight ambitious builds, it produced unusually complete games, interfaces, music, and CAD files for a creator-reported $0.79 of DeepSeek API spend.

The same test also found a stable weakness profile: inverted controls, wrong 3D orientation, mangled vehicles, fragile game state, missing enemies, and generated objects that looked plausible before close inspection. This is a strong prototyping worker. It is not a one-prompt production engineer, a reliable CAD authority, or a model to run unattended against important systems.

JQ AI SYSTEMS verdict: use V4 Flash for high-volume first passes, visual prototypes, disposable test code, repository exploration, and parallel agent work where every result has a verifier. Use the API first. Consider self-hosting only when data control justifies specialist hardware and slower low-bit inference.

Watch the Full Test

Video credit: Bijan Bowen. Watch the original video on YouTube. Model architecture, license, benchmark settings, API availability, and pricing were checked against DeepSeek's official model card and API documentation on 3 August 2026. Visual quality judgments and the $0.79 spend are creator observations from this run.

What DeepSeek Actually Released

DeepSeek-V4-Flash-0731 is the official release that supersedes the preview. DeepSeek says the architecture and size are unchanged and that the improvement comes from additional post-training. The release also attaches DeepSeek's DSpark speculative-decoding module, supports low, high, and max reasoning effort, and is adapted to the Responses API used by Codex.

SpecificationOfficial valueWhat it means
ArchitectureMixture of ExpertsOnly part of the network is active for each token.
Total parameters284BWeights and memory remain very large even though active compute is lower.
Activated parameters13BThe source of the “small” and low-cost framing.
Context1M tokensUseful for large repositories and long agent traces; full context still needs substantial KV-cache memory.
Maximum output384K tokensA ceiling, not a sensible default budget.
Reasoning effortLow, high, maxBijan tested at max, favoring capability over speed and cost.
WeightsMIT LicenseCommercial reuse and modification are permitted under the license terms.
Hosted accessPublic beta APIUse deepseek-v4-flash for the current 0731 endpoint.

Is It Really a Small Model?

Calling a 284B model “small” is reasonable only inside the narrow world of frontier Mixture-of-Experts systems. V4 Flash activates 13B parameters for a token, far below DeepSeek V4 Pro's 49B active path and Kimi K3's much larger total footprint. That reduces inference work. It does not shrink every weight, the context cache, runtime buffers, or the speculative module into a normal laptop package.

A better description is small active path, large physical model. A 13B dense model may fit comfortably on a consumer machine. A 284B MoE still needs all experts available in memory or across devices, even though only selected experts execute for each token. The distinction matters before anyone buys hardware based on the headline.

The Real API Price

The video reads the cache-hit figure as roughly three cents. DeepSeek's pricing table actually lists $0.0028 per million cache-hit input tokens: twenty-eight hundredths of one cent. The cache-miss and output figures in the video are correct. DeepSeek also announces a future peak-hours multiplier, but the same page says the effective date will follow in a separate announcement.

Billing itemRegular price per 1M tokensFuture peak-hours price
Input, cache hit$0.0028$0.0056
Input, cache miss$0.14$0.28
Output$0.28$0.56

During the video, the displayed DeepSeek balance moves from $9.85 to $9.06, implying $0.79 for the model work shown. The Browser OS alone moves the balance from $9.85 to $9.77, about eight cents. These numbers are useful directional evidence, not a controlled cost benchmark: tasks ran with different lengths, one game ran concurrently, cache status was not reported, and the Steve's PC Repair build used a separate OpenAI API key for dialogue.

The practical buying metric is not dollars per million tokens. It is cost per accepted result. A $0.10 build that needs an hour of debugging may be more expensive than a $2 build that passes its tests. Track model spend, human repair time, retries, and verifier cost together.

The Eight-Test Scorecard

TestStrongest signalObserved failureEvidence confidence
Browser OSResizable windows, files, right-click menu, personalization, games, snapshots, local date and timeCalculator error, incomplete settings, inverted or mismapped game controlsMedium: rich result, but the prompt is a recurring public benchmark
C++ skate gameComplete retro scene, pedestrians, tricks, rail balance, beach and water interactionBroken text and tricks on first pass; awkward HUD after repairMedium-high: improved through visible iteration
Subway FPSAtmosphere, sound, bloom toggle, flashlight, bullet impacts, exploration and combat modesNo enemies initially, auto-firing weapon, unreliable wave logicHigh for visual prototyping; low for game correctness
C++ MotoGPStrong retro racing presentation in both runsFirst run violated the no-raylib constraint; second may have retained prior context; controls appeared invertedLow: Bijan correctly calls the test ruined
Luxury watch siteGood materials, terminology, rendered assets, sub-dials, and product-page structureWatch parts were arranged or oriented incorrectly in 3DMedium: good style signal, poor product fidelity
Printable inline-six shellMultiple STL parts, shaft opening, vents, support-aware geometry, motor fit-check modelMotor orientation wrong, some walls too thin, uncertain ring and cover fitMedium: promising CAD draft, no physical print test
C++ Chrono CityFive eras, changing architecture, signs, palette, music, people, and cinematic transitions without raylibLong black-screen debugging, mangled cars, odd movement, hard-to-control tourMedium-high for ambition; medium for usability
Steve's PC RepairLong narrative, repair interaction, dialogue, inventory, travel, stealth scene, and multi-stage progressionUnclear navigation, pacing too fast, spatial confusion, progression became frustratingMedium: impressive scope, incomplete playability

What V4 Flash Did Well

The strongest pattern was not perfect code. It was scope completion. V4 Flash repeatedly turned broad prompts into working folders with application structure, rendering, sound, assets, interactions, and an attempt at verification. That is valuable for builders who need ten plausible directions before choosing one.

  1. End-to-end prototyping: it built playable scenes rather than stopping at scaffolding or an implementation plan.
  2. Multimodal-feeling craft from text: materials, music, lighting, UI, and motion often formed a coherent mood even when geometry was wrong.
  3. Iteration after feedback: the skate game recovered an ollie, tricks, and some text after a short correction loop.
  4. Tool endurance: the model kept working through compile, browser, rendering, and file-generation tasks under max reasoning.
  5. Novel combinations: C++ Chrono City and the inline-six motor shell pushed beyond familiar landing-page tests.
  6. Economics: the reported endpoint spend is low enough to run several candidates and spend the saved budget on verification.

Where It Failed, Repeatedly

A good evaluation looks for recurring failure modes, not a gallery of best screenshots. Across unrelated prompts, V4 Flash showed the same weaknesses often enough to treat them as routing information.

  1. Spatial reasoning: 3D watch parts, motor placement, vehicle orientation, and camera controls were often wrong while still looking superficially polished.
  2. Control mapping: inverted movement and unclear key bindings appeared in the Browser OS games, MotoGP, Chrono City, and PC Repair.
  3. State and game logic: missing enemies, wave zero, auto-fire, unclear task progression, and over-fast dialogue reduced playability.
  4. Constraint obedience: the first MotoGP run used raylib despite an explicit prohibition.
  5. Verification depth: rendered screenshots and plausible geometry are not proof that a part fits, a game loop completes, or a calculation is correct.
Routing rule: let V4 Flash generate the candidate; make deterministic tools judge it. Use compiler warnings, unit and integration tests, browser assertions, collision tests, mesh checks, dimensional constraints, screenshots, and a second model review before accepting the result.

Benchmark Context: Strong, Not an Overall Winner

DeepSeek reports a large jump from the preview after post-training. On the public tests listed in the official release, V4 Flash 0731 scores 82.7 on Terminal-Bench 2.1, 54.2 on NL2Repo, 76.7 on Cybergym, 54.4 on DeepSWE, and 70.3 on Toolathlon-Verified. It beats DeepSeek V4 Pro Preview across DeepSeek's table, despite the smaller active path.

It does not beat Opus 4.8 across that same table. Opus is higher on every listed public benchmark: 85.0 on Terminal-Bench, 69.7 on NL2Repo, 83.1 on Cybergym, 58.0 on DeepSWE, and 76.2 on Toolathlon-Verified. DeepSeek's two DSBench results are internal. The code-agent runs also use a forthcoming DeepSeek Harness in minimal mode, max effort, temperature=1.0, and top_p=0.95.

That makes “best small model” a reasonable creator verdict, not a settled benchmark fact. A useful comparison would run V4 Flash, Kimi K3, GPT-5.6, and Claude on the same repository snapshot, same acceptance tests, same tool permissions, same retry budget, and fresh sessions. Bijan's video is a strong capability probe; it is not that controlled experiment.

Local Hardware Reality

DeepSeek's official mixed-precision checkpoint uses FP4 expert weights and FP8 for most other parameters. Its published vLLM example serves the 0731 release on a single four-GPU GB300 node. That is the clean reference deployment, and it is data-center hardware.

The video points to a community four-bit quant around 138 GB and more aggressive dynamic quants that might squeeze into a 128 GB unified-memory machine. “Might fit” needs five caveats: the quant is not the official checkpoint format, quality may fall, runtime overhead also needs memory, a long context can make the KV cache enormous, and token speed may be too slow for interactive agent work. Two DGX Spark systems are an enthusiast route, not a consumer baseline or an official requirement.

RouteBest forMain tradeoff
DeepSeek APIMost developers, evaluation, bursty agent workHosted-data policy, network dependence, changing prices
Third-party hosted endpointRegion or infrastructure choiceProvider quality, terms, and model-version drift
Official mixed-precision self-hostLabs and companies with multi-GPU infrastructureHardware, power, orchestration, and operating cost
Community low-bit quantLocal AI enthusiasts exploring fit and privacyQuality, speed, context, tooling, and reproducibility
Smaller dense local modelDaily private work on ordinary hardwareLower peak capability, often much better responsiveness

Using DeepSeek V4 Flash in Codex

DeepSeek publishes an official Codex integration guide. V4 Flash supports the Responses API, and DeepSeek says one Codex configuration can expose the model to the CLI, desktop app, and IDE extension. The guide offers an automated setup script and a manual configuration path.

Prefer the manual path, or download and inspect the script before execution. A command that pipes a remote script directly into a shell gives the publisher immediate code execution on your machine. The official script says it backs up ~/.codex/config.toml, writes a model catalog, preserves MCP and trust settings, and validates files before writing; inspecting the current script is still the safer habit.

  1. Create a separate DeepSeek API key with a small prepaid balance.
  2. Back up the Codex configuration and inspect the provider changes.
  3. Open a fresh, non-sensitive repository with no production credentials.
  4. Start at high effort; reserve max for genuinely difficult tasks.
  5. Give the task a written acceptance test and a hard cost or time limit.
  6. Review the diff and run the repository's full verification before merging.

A Better Seven-Step Evaluation

  1. Choose one real task. Use a bug, feature, migration, or design repair from your own work rather than a famous public benchmark.
  2. Freeze the environment. Same commit, dependencies, tools, permissions, and machine for every model.
  3. Write acceptance criteria first. Include tests, screenshot states, performance limits, accessibility checks, or dimensional tolerances.
  4. Start fresh sessions. This prevents prior attempts from leaking hints into later runs.
  5. Cap retries. Give each model the same time, tokens, and number of repair loops.
  6. Score the whole outcome. Track pass rate, latency, API cost, human repair minutes, and regressions.
  7. Route by failure profile. Keep V4 Flash for the tasks where it wins on accepted-result cost, not where it only makes the best first screenshot.

Data and Privacy Before Hosted Use

DeepSeek's current public privacy policy says its services may collect prompts, uploaded files, feedback, chat history, and generated outputs; may use data to improve and train its technology; offers a training opt-out; and processes and stores personal data in the People's Republic of China. It also says the services are not designed for sensitive personal data.

That does not automatically disqualify the API. It does mean a cheap model should not bypass normal vendor review. Keep secrets out of prompts, redact client data, use synthetic fixtures, restrict agent permissions, rotate keys, and obtain legal or security approval before sending confidential repositories. Self-hosting the MIT weights can change the data path, but then patching, access control, logging, and infrastructure security become your responsibility.

Claim Audit

ClaimAssessmentContext
V4 Flash is the best small model yet.Opinion, plausibleExcellent creator results and low cost; no controlled cross-model test establishes a universal winner.
The model is 284B total and 13B active.ConfirmedPublished in DeepSeek's official V4 model card; the 0731 release keeps the same architecture.
It has a one-million-token context.ConfirmedListed in the official model card and API pricing documentation.
Cache hits cost about three cents per million input tokens.IncorrectThe official figure is $0.0028, about 0.28 cents per million.
All shown tests cost $0.79.Creator-reportedThe displayed DeepSeek balance falls by $0.79; a separate OpenAI dialogue cost and human time are excluded.
It can run on one 128 GB machine.Possible only with caveatsAn aggressive community quant may fit, but speed, overhead, context, quality, and runtime support determine practicality.
V4 Flash 0731 beats V4 Pro Preview.Project-reportedIt leads across DeepSeek's published table after additional post-training.
The future V4 Pro will beat Kimi K3.SpeculationFlash improvements do not guarantee how the unreleased Pro update will perform.
The model is safe to ship from one prompt.FalseThe tests show repeated logic, geometry, control, and constraint failures despite polished presentation.

Video Chapters

TimeChapterEvaluation signal
00:00IntroOfficial 0731 release and the “small model” thesis.
01:11Local AI info284B total, 13B active, community quantization estimates.
02:45Technical lookBenchmark jump, API access, context, and pricing.
04:26Browser OS testRich interactivity plus arithmetic and control failures.
11:11C++ skate gameStrong scene construction and useful iteration after feedback.
14:13Subway FPSExcellent atmosphere; missing enemies and broken combat state.
16:50C++ MotoGPVisually strong but methodologically contaminated.
18:52Watch websiteGood materials and copy; weak 3D orientation.
20:473D CAD modelPrintable-looking parts with dimensional and fit concerns.
23:35C++ Chrono CityNovel multi-era build with music, motion, and spatial defects.
28:04Total API costCreator-reported $0.79 DeepSeek balance change.
28:23Steve's PC RepairLong-form narrative ambition, but unreliable navigation and pacing.
34:04Results overviewStrength and failure patterns across all tests.
36:23Closing thoughtsLocal-hosting enthusiasm and unverified V4 Pro forecast.

Bottom Line

DeepSeek V4 Flash 0731 is not impressive because it made one pretty webpage. It is impressive because a 13B-active open-weight model repeatedly assembled ambitious, multi-file, interactive projects at a remarkably low hosted price. Its official agent benchmarks also show that post-training moved it far beyond the preview.

The video is equally useful as a warning. Polished outputs concealed arithmetic mistakes, wrong orientations, violated constraints, broken controls, and incomplete state logic. Those are exactly the errors that become expensive when a prototype is mistaken for a finished system.

The sensible adoption path is simple: test the API on a real low-risk task, give every task an objective verifier, measure accepted-result cost, and route work by observed strengths. Keep local V4 Flash as an infrastructure decision, not an impulse hardware purchase. Under those conditions, this model can be a very capable and unusually affordable worker.

Sources and Useful Links

Common questions

Is DeepSeek V4 Flash a small model?
Only in a relative compute sense. It is a Mixture-of-Experts model with 284 billion total parameters and about 13 billion activated per token. That active path helps serving efficiency, but the full weights and runtime overhead are far beyond an ordinary laptop model.
Is DeepSeek V4 Flash open source?
The official DeepSeek-V4-Flash-0731 repository publishes the model weights under the MIT License. Open weights do not make deployment easy: the official serving example uses a four-GPU GB300 node, while smaller community quantizations trade quality and context capacity for memory savings.
How much does the DeepSeek V4 Flash API cost?
As checked on 3 August 2026, DeepSeek lists V4 Flash at $0.0028 per million cache-hit input tokens, $0.14 per million cache-miss input tokens, and $0.28 per million output tokens. DeepSeek says a future peak-hours policy will double those prices, but its pricing page says the effective date will be announced separately.
Did Bijan Bowen spend only $0.79 on all eight tests?
The video shows the DeepSeek balance moving from $9.85 to $9.06, a reported $0.79 for the model work shown, including a long-running game task. That is a creator-observed endpoint cost, not a controlled billing benchmark. The game also used a separate OpenAI API key for dialogue, whose cost was not included.
What is DeepSeek V4 Flash best at?
The strongest evidence in this test is rapid end-to-end prototyping: interactive interfaces, C++ game scenes, generated music and assets, multi-file CAD output, and iterative repairs. The model repeatedly struggled with 3D orientation, control mapping, state logic, and polished one-shot correctness.
Can DeepSeek V4 Flash run on a 128 GB Mac or workstation?
A very aggressive community quantization may fit close to that memory envelope, but fitting weights is not the same as a practical deployment. Runtime overhead, KV cache, context length, speed, quantization loss, and software support all matter. The official mixed-precision checkpoint is aimed at multi-GPU infrastructure.
Can DeepSeek V4 Flash be used in Codex?
Yes. DeepSeek V4 Flash supports the Responses API and DeepSeek publishes an official Codex integration guide. Review any setup script before running it, back up the Codex configuration, use a fresh low-risk repository, and keep secrets and production credentials out of the first test.
Share
X LinkedIn Reddit
Build Yours

Want a system
like this one?

Book a free 30-minute call. We map your situation, identify the highest-impact automation, and figure out if we are a fit.

Book Free 30-min Call