deepseek-v4.1-flashDeepSeek
- endpoints
- 12
- regions
- 2
- best tok/s
- 320
- api
- chat
- released
- September 2026
- eu routing
- available
capabilities 6/8
The same weights cost 3.3x more on lyceum than on runware. That is what the endpoints table is for.
| # | flags | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| 1 |  Runware | Global | 1.0M | $0.15 | $0.60 | $0.01 | 1.67s | 75.0 | zdr | |
| 2 |  DeepInfra | Global | 1.0M | $0.16 | $0.48 | $0.0048 | 1.49s | 61.0 | - | zdr |
| 3 | DeepInfra | Global | 1.0M | $0.20 | $0.60 | $0.0060 | 1.49s | 61.0 | - | zdr |
| 4 |  Fireworks AI | Global | 1.0M | $0.22 | $0.66 | $0.0070 | 1.29s | 61.0 | - | zdr |
| 5 |  Melious | EU | 1M | $0.23 | $1.14 | $0.02 | 775ms | 69.0 | zdr | |
| 6 |  DeepSeek | Global | 1M | $0.30 | $1.20 | $0.0060 | 1.20s | 122.0 | - | no-train |
| 7 |  Novita AI | Global | 1.0M | $0.30 | $1.20 | $0.0060 | 2.92s | 94.0 | - | zdr |
| 8 |  Tencent | Global | 1.0M | $0.30 | $1.20 | $0.0060 | 1.78s | 24.0 | - | no-train |
| 9 |  Together AI | Global | 1.0M | $0.30 | $1.20 | $0.01 | 338ms | 227.0 | - | zdr |
| 10 |  Lyceum | EU | 1.0M | $0.50 | $1.50 | $0.13 | 1.59s | 267.0 | zdr | |
| 11 |  sference | EU | 1.0M | $0.50 | $1.50 | $0.05 | 562ms | 243.0 | zdr | |
| 12 |  TensorX | EU | 1.0M | $0.50 | $1.50 | $0.13 | 2.03s | 320.0 | zdr |
| provider | now | low | high | change |
|---|---|---|---|---|
| Together AI | 338ms | 370ms | 2.80s | +381% |
| sference | 562ms | 71ms | 794ms | +955% |
| Melious | 775ms | 608ms | 689ms | +7.2% |
| DeepSeek | 1.20s | 752ms | 1.50s | +70% |
| Fireworks AI | 1.29s | 1.19s | 2.31s | +71% |
| DeepInfra | 1.49s | 1.05s | 1.93s | +36% |
| Lyceum | 1.59s | 627ms | 2.50s | +300% |
| Runware | 1.67s | 1.60s | 2.10s | -5.4% |
| Tencent | 1.78s | 1.71s | 5.87s | -69% |
| TensorX | 2.03s | 1.00s | 1.48s | +19% |
| Novita AI | 2.92s | 2.08s | 4.08s | -4.6% |
median wait before the first token lands. this is the number a user feels, and it moves with provider load through the day.
Lines are steps: each sample is one week of traffic held flat, not a slide from the week before. Click a name in the key to drop it from the plot. Measured through Nov 2.
no published benchmark scores for this model
deepseek-v4.1-flashrequesty routes this id across every provider serving deepseek-v4.1-flash, picking on price and health and failing over automatically. the id stays valid when a provider changes underneath it
Base url is https://router.requesty.ai/v1 for every id here. One key reaches the whole catalog.
from openai import OpenAI client = OpenAI( api_key="YOUR_REQUESTY_API_KEY", base_url="https://router.requesty.ai/v1",) response = client.chat.completions.create( model="deepseek-v4.1-flash", messages=[ {"role": "user", "content": "Explain quantum computing in one paragraph."}, ],) print(response.choices[0].message.content)
Change the base url, use your Requesty key, set the model to any id in the call pane. Existing OpenAI SDK code needs no other edit, and the same key reaches every other model in the catalog, so switching later is a one-parameter change. Browse all models
DeepSeek V4.1 Flash is a 552B parameter Mixture of Experts model built on a new Causal Encoder Decoder architecture that activates only 8B parameters for input and 16B for output, giving frontier level intelligence at a fraction of the cost. It is the smallest model in the new DeepSeek architecture family and the first Flash with native vision and multimodal understanding. V4.1 Flash beats DeepSeek V4 Pro on performance, cost, speed and task completion time and scores 88.1 on CyberGym. Its KV cache is compressed to 890 bytes per token, four times smaller than V4 Flash and over 400 times smaller than V1, cutting the cache hit costs that dominate agent workloads. 1M token context window, 384K max output, thinking and non thinking modes, tool calling, JSON output and prompt caching. Pricing per 1M tokens starts at $0.003 for cache hits, $0.15 input and $0.60 output off peak, with peak rates at double (01:00 to 04:00 and 06:00 to 10:00 UTC, Monday to Friday). Open weights are published on Hugging Face. Ideal for coding agents, long context RAG, high throughput batch processing and multimodal assistants. Replaces deepseek-v4-flash and deepseek-v4-flash-vision-exp, both still accepted as aliases.
questions 7
Which providers serve deepseek-v4.1-flash?
How much does deepseek-v4.1-flash cost?
Is deepseek-v4.1-flash open weights?
What is the context window of deepseek-v4.1-flash?
Which provider is fastest for deepseek-v4.1-flash?
How do I use deepseek-v4.1-flash with the OpenAI SDK?
Should I call deepseek-v4.1-flash by its managed id or a provider id?
more from deepseek 6
| model | context | from /M | providers |
|---|---|---|---|
| deepseek-v4.1-flash-1017 | 1.0M | $0.10 | 8 |
| deepseek-v4-flash-vision-exp | 1.0M | $0.22 | 1 |
| deepseek-v4-pro-0813 | 1.0M | $1.30 | 5 |
| deepseek-v4-flash-0731 | 1.0M | $0.07 | 10 |
| deepseek-v4-pro-0424 | 1.0M | $1.04 | 4 |
| deepseek-v4-flash-0424 | 1.0M | $0.07 | 4 |
route deepseek-v4.1-flash through one endpoint
One key for 11 providers on this model and 600+ others. No markup on provider prices, automatic failover, caching built in. Weekly aggregates in the measured pane come from production traffic routed through Requesty, one line per provider on a shared axis. Methodology