DeepSeek V4.1 Flash
deepseek/deepseek-v4.1-flashDeepSeek V4.1 Flash is a sparse mixture-of-experts model from DeepSeek, and the first built on the company's Causal Encoder-Decoder (CED) architecture.
Cheap mode · 0.6–64.2% off ↓Cheap mode · 0.6–64.2% off
Fast mode prioritizes response speed. Cheap mode uses lower-cost capacity and may be a little slower. Both run DeepSeek V4.1 Flash.
Per 1M tokens, including your 5% fee. Your selection carries into the playground and API example.
| Per 1 million tokens | Fast mode | Cheap mode | You save |
|---|---|---|---|
| Input | $0.0536 | $0.0532 | 0.6% |
| Output | $0.5954 | $0.2129 | 64.2% |
All displayed prices include the 5% Free-plan fee. Standard rates are OpenRouter’s published routing ceilings; the actual standard bill may be lower. Cheapest quote: Baidu. Savings compare uncached token rates; other provider charges and your usage mix can change the total. Availability is not guaranteed.
Quote checked 11/8/2026, 8:18:04 PM UTC.
Choose how to use DeepSeek V4.1 Flash
Playground
Try prompts in your browser, create an agent, and see the cost of each reply. Your selected model carries into setup.
Model capabilities below describe the underlying model. Available playground features depend on its provider.
API
Connect your app with a Gizmo API key. Choose fastest or cheapest, stream text responses, and inspect timing, tokens, and charges in request logs.
A current API route is listed for this exact model. Gateway configuration is required for live calls.
Provider price + a transparent platform fee
| Rate | Provider baseline | Free plan · 5% fee | Paid plan · 4% fee |
|---|---|---|---|
| Input / 1M tokens | $0.051 | $0.05355 | $0.05304 |
| Output / 1M tokens | $0.567 | $0.59535 | $0.58968 |
| API input / 1M tokens | $0.051 | $0.05355 | $0.05304 |
| API output / 1M tokens | $0.567 | $0.59535 | $0.58968 |
Legacy chat image rates use the catalog baseline. The dedicated image playground shows separate estimates and settles the provider’s metered usage. API rates are current routing ceilings; final API charges use provider-reported cost plus your fee. Cheapest routing compares eligible OpenRouter and Cheaper Inference rates for the same model; savings are not guaranteed. Cached input or other provider charges may differ. Compare all model rates and verified discounts →
How your plan pays for usage
API and playground share one included allowance. Usage including the fee draws from that allowance first, then your prepaid credits. Plans bill monthly; included usage resets weekly: Free $1, Starter $5, Growth $25, Scale $50.
A $1 provider charge becomes $1.05 on Free or $1.04 on a paid plan before included usage is applied. Plan eligibility still applies in the playground.
Free models and limits
Zero-cost models stay $0, including the fee, and do not consume your dollar allowance. Free accounts can make 1,000 free-model API requests per calendar month, resetting in UTC. Admitted requests count even if the provider fails.
Free includes 20 API requests/minute and 6 playground messages/minute. Paid plans raise these limits. See plan limits.
Before a text API request, Gizmo temporarily reserves included usage and, if necessary, prepaid credit for the model’s full context plus your maximum output. Image jobs reserve twice their displayed estimate. Unused funds return after metering. This hold can exceed the final charge; unconfirmed usage remains marked pending.
Overview
DeepSeek V4.1 Flash is a sparse mixture-of-experts model from DeepSeek, and the first built on the company's Causal Encoder-Decoder (CED) architecture. It activates 8B parameters on input and 16B on output from a 552B-parameter backbone, an asymmetric split that keeps per-token compute low relative to the model's total size. Image understanding is native to the architecture, with visual and text embeddings trained jointly from the start of pre-training rather than added afterward as in the earlier experimental V4 Flash Vision Exp(opens in new tab).
It is suited for coding, terminal, and computer-use agents, along with long-horizon tasks that must run to completion across many steps and long-context document and chart analysis. Compressed KV caching cuts cache memory to roughly a quarter of the previous Flash generation, which matters for agent loops that repeatedly re-read large contexts. DeepSeek positions it as the cost-efficient tier of the V4.1 family and reports that it exceeds V4 Pro(opens in new tab) on performance, speed, and task completion time.
Benchmarks
Base Model from the model card
| Benchmark (Metric) | # Shots | DeepSeek-V4-Flash-Base | DeepSeek-V4-Pro-Base | DeepSeek-V4.1-Flash-Base |
|---|---|---|---|---|
| Architecture | — | MoE | MoE | MoE |
| # Backbone Params | — | 284B | 1.6T | 552B |
| # Activated Params | — | 13B | 49B | 8B / 16B |
| World Knowledge | ||||
| AGIEval (EM) | 3–5-shot | 83.9 | 84.4 | 83.4 |
| MMLU-Pro (EM) | 5-shot | 68.3 | 73.5 | 74.1 |
| C-Eval (EM) | 5-shot | 92.1 | 93.1 | 92.1 |
| MultiLoKo (LLM-Judge) | 5-shot | 42.6 | 50.9 | 45.5 |
| SimpleQA-Verified (EM) | 25-shot | 30.1 | 55.2 | 42.3 |
| SuperGPQA (EM) | 5-shot | 46.5 | 53.9 | 53.1 |
| Language & Reasoning | ||||
| BBH (EM) | 3-shot | 86.9 | 87.5 | 86.1 |
| BBEH (EM) | 1-shot | 25.4 | 29.8 | 27.2 |
| DROP (F1) | 1-shot | 88.6 | 88.7 | 87.9 |
| HellaSwag (EM) | 0-shot | 85.7 | 88.0 | 87.2 |
| Code & Math | ||||
| BigCodeBench (Pass@1) | 3-shot | 56.8 | 59.2 | 60.6 |
| HumanEval (Pass@1) | 0-shot | 69.5 | 76.8 | 79.4 |
| GSM8K (EM) | 8-shot | 90.8 | 92.6 | 93.0 |
| MATH (EM) | 4-shot | 57.4 | 64.5 | 61.1 |
| MGSM (EM) | 8-shot | 85.7 | 84.4 | 80.2 |
| Long Context | ||||
| LongBench-V2 (EM) | 1-shot | 44.7 | 51.5 | 45.2 |
| Multimodal | ||||
Comparison with frontier models (Max reasoning effort) from the model card
| Benchmark (Metric) | Opus-5.0 | GPT-5.6 Sol | K3 | GLM-5.3 | DS-V4-Pro | DS-V4-Flash | DS-V4.1-Flash |
|---|---|---|---|---|---|---|---|
| Reasoning | |||||||
| GPQA Diamond (Pass@1) | 93.4 | 94.1 | 92.9 | 88.1 | 92.4 | 89.9 | 90.9 |
| HLE (Pass@1) | 56.3 | 44.5 | 43.5 | 42.0† | 42.7† | 37.8† | 36.8 (39.1†) |
| Codeforces (Rating) | — | — | — | — | 3348 | 3289 | 3471 |
| MathArena Apex (Pass@1) | — | — | 65.6 | — | 65.3 | 58.6 | 65.6 |
| Agentic | |||||||
| Terminal-Bench 2.1 (Pass@1) | 89.1 | 88.8 | 88.3 | 88.2 | 87.9 | 82.7 | 90.6 |
| Terminal-Bench 3.0 (Pass@1) | 43.3 | 34.4 | 17.7 | 28.3 | 11.8 | 7.6 | 30.0 |
| Terminal-Bench 4.0 (Pass@1) | 51.8 | 39.9 | 12.6 | 37.9 | 12.4 | 7.0 | 31.2 |
| DeepSWE v1.1 (Resolved) | 74.0 | 73.0 | 67.5 | 66.9 | 62.7 | 54.4 | 74.2 |
| ProgramBench (Almost@1) | 37.0 | 23.0 | 17.5 | 19.0 | 15.5 | — | 20.3 |
| NL2Repo-Bench (Score) | 75.3 | 56.8 | 58.0 | 58.0 | 61.5 | 54.2 | 64.0 |
| CyberGym (Pass@1) | — | 84.5 | 80.0 | 84.5 | 83.3 | 76.7 | 88.1 |
| SEC-Bench Pro (Pass@1) | — | 74.3 | — | — | 56.4 | 30.9 | 62.8 |
| ExploitGym (Pass@1) | 22.1 | 33.7 | — | 15.0 | 5.4 | 1.8 | 15.3 |
| HLE w/ tools (Pass@1) | 63.6 | — | 59.8 | 62.5 | 60.0 | 51.5 | 63.9 |
| AutomationBench (Pass@1) | 50.3 | 45.8 | 46.7 | 48.8 | 43.2 | 37.7 | 54.8 |
| Agent's Last Exam (Pass@1) | 28.6 | 26.7 | 27.6 | 28.5 | 25.7 | 25.2 | 31.8 |
| Chartography w/ tools (Pass@1) | 84.0 | 79.9 | 68.1 | — | — | — | 78.9 |
| BabyVision w/ tools (Pass@1) | 94.1 | 88.9 | 85.7 | — | — | — | 89.6 |
| ZeroBench-main w/ tools (Pass@5) | 52.0 | 53.0 | 41.0 | — | — | — | 49.0 |
Performance across agent scaffolds (DeepSWE v1.1 and Terminal-Bench 2.1, Max reasoning effort) from the model card
| Benchmark (Metric) | Claude Code | Codex | OpenCode | Pi | mini-SWE | DSH Minimal | DSH Standard |
|---|---|---|---|---|---|---|---|
| DeepSWE v1.1 (Resolved) | 69.8 | 65.6 | 65.5 | 66.2 | 74.2 | 72.6 | 70.5 |
| Terminal-Bench 2.1 (Pass@1) | 88.0 | 84.1 | 85.0 | 86.1 | 90.3 | 90.6 | 85.8 |
Capabilities
Reasoning is on by default with effort levels max · high · low (default high).
Sampling controls: temperature, top_p, top_k, min_p, max_tokens, frequency_penalty, presence_penalty, repetition_penalty.
Pricing on Gizmo
These are provider baseline rates before Gizmo’s 5% Free-plan or 4% paid-plan fee. Your included usage covers charges first; overage uses prepaid credits. See the fee-inclusive comparison above.
Provider baseline rates
- Input
- $0.051 / 1M
- Cached input
- $0.003 / 1M
- Output
- $0.567 / 1M
Examples before the platform fee
- Typical chat turn
- $0.00039 2k in · 500 out
- $1 of input
- ≈ 19.61M tokens
- $1 of output
- ≈ 1.76M tokens
- Full context, once
- $0.0535
Providers and pricing
4 ways to run DeepSeek V4.1 Flash on Gizmo, cheapest first. Requests follow your routing choice and fall back down this list if a provider fails.
| Provider | Route | Input / 1M | Output / 1M | Context | Uptime 24h |
|---|---|---|---|---|---|
| Baidu · fp8Lowest price | OpenRouter | $0.0507 | $0.20 | 1.05M | 100% |
| Morph · fp8 | OpenRouter | $0.022 | $0.40 | 1.05M | 100% |
| Relace · fp4 | OpenRouter | $0.006 | $0.50 | 1.05M | 98.7% |
| Wafer · us | OpenRouter | $0.051 | $0.57 | 1.05M | 100% |
Rates are each provider’s, per million tokens, before Gizmo’s 5% (Free) or 4% (paid) fee. “Cheapest” routing starts at the top of this list; “Fastest” starts with the model’s maker.
More from DeepSeek
DeepSeek V4.1 Flash 1017DeepSeek · 1.05M ctx$0.006 · $0.5DeepSeek V4.1 Flash 1017 (batch)DeepSeek · 1.05M ctx$0.011 · $0.2DeepSeek V4.1 Flash (batch)DeepSeek · 1.05M ctx$0.011 · $0.2DeepSeek V4 Flash Vision ExpDeepSeek · 1.05M ctx$0.44 · $1.32DeepSeek V4 Pro 0813DeepSeek · 1.05M ctx$0.111 · $4.44DeepSeek V4 Flash 0731DeepSeek · 1.05M ctx$0.043 · $0.3DeepSeek V4 Flash 0731 (batch)DeepSeek · 1.05M ctx$0.048 · $0.144DeepSeek V4 Pro 0423DeepSeek · 1.05M ctx$0.209 · $0.418Similar price, other creators
Run DeepSeek V4.1 Flash in under a minute
Create your account
Sign up with just an email. Start on the free tier with no card, or pick a paid plan for a bigger weekly allowance and $25 in bonus credits.
Your agent is ready
DeepSeek V4.1 Flash is preselected as the brain of your first agent. Give it a name and a system prompt, or keep the defaults.
Chat, research, create
Live web research and image generation are built in. Every reply shows what it cost, and you can switch models anytime.