Where is the DeepSeek V4.1 Flash 1017 documentation?

The DeepSeek V4.1 Flash 1017 documentation on OpenRouter includes the Best OCR Model for Scanned Documents, Measured blog post, the Build an AI Code Reviewer with OpenRouter blog post, the Open-Weight vs Closed-Weight Models: What's the Difference? blog post, the How to Extract Tables from a PDF with an LLM blog post, and the Does DeepSeek V4 Have Vision? blog post. The quickstart covers API keys and your first request, and the API reference lists every parameter.

How much does DeepSeek V4.1 Flash 1017 cost?

DeepSeek V4.1 Flash 1017 costs $0.038/M input tokens and $0.50/M output tokens, with separate rates for Cache Read at $0.01/M tokens.

What is the context length of DeepSeek V4.1 Flash 1017?

DeepSeek V4.1 Flash 1017 has a 1,048,576 token context window. It supports up to 1,048,576 completion tokens.

Does DeepSeek V4.1 Flash 1017 support tool calling and structured outputs?

Yes. DeepSeek V4.1 Flash 1017 accepts tools and tool_choice for function calling. It also supports structured outputs via a JSON schema in response_format.

What inputs and outputs does DeepSeek V4.1 Flash 1017 support?

DeepSeek V4.1 Flash 1017 accepts text and images as input and returns text.

Which providers serve DeepSeek V4.1 Flash 1017?

DeepSeek V4.1 Flash 1017 is served by 38 providers on OpenRouter: Particle AI, Wafer, Prime Intellect, Open Inference, Sail Research, Ionstream, DeepInfra, inference.net and 30 more. Requests are routed to the best available provider, with automatic failover to the others, and you can pin or exclude providers with provider routing.

When was DeepSeek V4.1 Flash 1017 released?

DeepSeek V4.1 Flash 1017 was released on October 17, 2026.

DeepSeek: DeepSeek V4.1 Flash 1017 (nitro)

deepseek/deepseek-v4.1-flash-1017:nitro

Model weights

Compare

DeepSeek V4.1 Flash 1017 is the October 17, 2026 snapshot of DeepSeek V4.1 Flash, served by DeepSeek under the deepseek-flash API model name. 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.

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 analysis. Compressed KV caching cuts cache memory to roughly a quarter of the previous Flash generation, significantly reducing costs on agentic workloads. DeepSeek positions it as the cost-efficient tier of the V4.1 family and reports that it exceeds V4 Pro on performance, speed, and task completion time.

Modalities

In / out price

$0.038 / $0.50per 1M

Context

1.0M

Released

Oct 17, 2026

Providers

Different companies host the same model. OpenRouter routes your request to one of them based on the routing mode you pick: Balanced (price + speed), Nitro (fastest), or Floor (cheapest).

Latency / throughput

Provider	Input /M	Output /M	Cache read /M	Latency	Throughput	Uptime
Fast	$0.60	$2.40	$0.014	0.32s	463 tok/s	99.43%
	$0.30	$1.20	$0.006	0.49s	209 tok/s	99.94%
	$0.30	$1.20	$0.007	0.85s	207 tok/s	99.67%
	$0.30	$1.20	$0.006	0.31s	200 tok/s	99.46%
	$0.29	$1.20	$0.007	0.41s	196 tok/s	100.00%
	$0.18	$0.72	$0.006	0.53s	184 tok/s	100.00%
	$0.45	$1.80	$0.009	0.61s	184 tok/s	99.97%
	$0.066	$0.50	$0.0065	0.45s	176 tok/s	100.00%
	$0.038	$0.50	$0.01	0.46s	175 tok/s	100.00%
	$0.24	$0.96	$0.0048	0.29s	172 tok/s	98.26%
	$0.21	$0.62	$0.0105	0.41s	162 tok/s	98.38%
	$0.30	$1.20	$0.006	0.78s	162 tok/s	99.90%
	$0.15	$0.50	$0.003	0.47s	158 tok/s	100.00%
	$0.12	$0.50	$0.003	0.59s	150 tok/s	99.97%
65% off	$0.30$0.105	$1.20$0.42	$0.006$0.0021	0.58s	148 tok/s	100.00%
	$0.15	$1.20	$0.006	0.92s	148 tok/s	99.93%
	$0.30	$1.20	$0.006	0.45s	147 tok/s	100.00%
	$0.16	$1.10	$0.008	0.41s	139 tok/s	99.72%
	$0.104	$0.583	$0.005	0.49s	133 tok/s	97.81%
40% off	$0.30$0.18	$1.20$0.72	$0.006$0.0036	0.71s	133 tok/s	99.65%
	$0.25	$1.00	$0.005	0.62s	131 tok/s	100.00%
	$0.30	$1.20	$0.007	0.63s	125 tok/s	100.00%
	$0.06	$0.50	$0.01	0.48s	124 tok/s	100.00%
	$0.08	$0.605	$0.008	0.22s	122 tok/s	100.00%
	$0.21	$0.84	$0.0042	0.74s	122 tok/s	98.50%
	$0.07	$0.80	$0.006	0.88s	120 tok/s	100.00%
	$0.04368	$0.70	$0.0024	0.84s	117 tok/s	99.97%
	$0.07272	$0.108	$0.007	1.41s	117 tok/s	99.76%
	$0.07272	$0.108	$0.007	1.38s	116 tok/s	99.78%
	$0.30	$1.20	$0.0075	0.84s	114 tok/s	100.00%
	$0.15	$0.60	$0.003	1.23s	114 tok/s	100.00%
65% off	$0.28$0.09772	$1.13$0.3944	$0.006$0.002094	0.69s	111 tok/s	99.10%
Fast	$0.104	$0.583	$0.005	0.57s	109 tok/s	100.00%
	$0.08	$0.605	$0.008	0.27s	102 tok/s	100.00%
	$0.21	$0.84	$0.0042	1.78s	100 tok/s	96.19%
	$0.12	$0.48	$0.0024	0.78s	99 tok/s	100.00%
51% off	$0.20$0.098	$0.60$0.294	$0.006$0.00294	0.97s	97 tok/s	99.91%
Priority51% off	$0.30$0.147	$0.90$0.441	$0.009$0.00441	0.96s	97 tok/s	99.97%
	$0.10	$0.60	$0.003	1.45s	93 tok/s	99.96%
	$0.18	$0.68	$0.006	0.85s	90 tok/s	99.17%
45% off	$0.30$0.165	$1.20$0.66	$0.006$0.0033	1.40s	89 tok/s	100.00%
	$0.30	$1.20	$0.006	1.88s	82 tok/s	99.80%
	$0.195	$0.78	$0.0039	2.66s	74 tok/s	99.96%
	$0.30	$0.55	$0.006	0.96s	71 tok/s	99.78%
63% off	$0.30$0.111	$1.20$0.444	$0.006$0.00222	1.90s	66 tok/s	100.00%
	$0.30	$1.20	$0.006	0.49s	40 tok/s	100.00%
Not used in Nitro routing:Why these endpoints are not used
Flex	$0.0775	$0.2325	$0.002325	0.66s	193 tok/s	99.94%
Flex	$0.03	$0.25	$0.005	0.50s	122 tok/s	100.00%
Flex51% off	$0.16$0.0784	$0.48$0.2352	$0.0048$0.002352	2.04s	86 tok/s	94.88%

Input $0.60Output $2.40
Fast
463 tok/s99.43% uptime
Input $0.30Output $1.20
209 tok/s99.94% uptime
Input $0.30Output $1.20
207 tok/s99.67% uptime
Input $0.30Output $1.20
200 tok/s99.46% uptime
Input $0.29Output $1.20
196 tok/s100.00% uptime
Input $0.18Output $0.72
184 tok/s100.00% uptime
Input $0.45Output $1.80
184 tok/s99.97% uptime
Input $0.066Output $0.50
176 tok/s100.00% uptime
Input $0.038Output $0.50
175 tok/s100.00% uptime
Input $0.24Output $0.96
172 tok/s98.26% uptime
Input $0.21Output $0.62
162 tok/s98.38% uptime
Input $0.30Output $1.20
162 tok/s99.90% uptime
Input $0.15Output $0.50
158 tok/s100.00% uptime
Input $0.12Output $0.50
150 tok/s99.97% uptime
Input $0.105Output $0.42
65% off
148 tok/s100.00% uptime
Input $0.15Output $1.20
148 tok/s99.93% uptime
Input $0.30Output $1.20
147 tok/s100.00% uptime
Input $0.16Output $1.10
139 tok/s99.72% uptime
Input $0.104Output $0.583
133 tok/s97.81% uptime
Input $0.18Output $0.72
40% off
133 tok/s99.65% uptime
Input $0.25Output $1.00
131 tok/s100.00% uptime
Input $0.30Output $1.20
125 tok/s100.00% uptime
Input $0.06Output $0.50
124 tok/s100.00% uptime
Input $0.08Output $0.605
122 tok/s100.00% uptime
Input $0.21Output $0.84
122 tok/s98.50% uptime
Input $0.07Output $0.80
120 tok/s100.00% uptime
Input $0.04368Output $0.70
117 tok/s99.97% uptime
Input $0.07272Output $0.108
117 tok/s99.76% uptime
Input $0.07272Output $0.108
116 tok/s99.78% uptime
Input $0.30Output $1.20
114 tok/s100.00% uptime
Input $0.15Output $0.60
114 tok/s100.00% uptime
Input $0.09772Output $0.3944
65% off
111 tok/s99.10% uptime
Input $0.104Output $0.583
Fast
109 tok/s100.00% uptime
Input $0.08Output $0.605
102 tok/s100.00% uptime
Input $0.21Output $0.84
100 tok/s96.19% uptime
Input $0.12Output $0.48
99 tok/s100.00% uptime
Input $0.098Output $0.294
51% off
97 tok/s99.91% uptime
Input $0.147Output $0.441
Priority
1 more label
97 tok/s99.97% uptime
Input $0.10Output $0.60
93 tok/s99.96% uptime
Input $0.18Output $0.68
90 tok/s99.17% uptime
Input $0.165Output $0.66
45% off
89 tok/s100.00% uptime
Input $0.30Output $1.20
82 tok/s99.80% uptime
Input $0.195Output $0.78
74 tok/s99.96% uptime
Input $0.30Output $0.55
71 tok/s99.78% uptime
Input $0.111Output $0.444
63% off
66 tok/s100.00% uptime
Input $0.30Output $1.20
40 tok/s100.00% uptime
Not used in Nitro routing:Why these endpoints are not used
Input $0.0775Output $0.2325
Flex
193 tok/s99.94% uptime
Input $0.03Output $0.25
Flex
122 tok/s100.00% uptime
Input $0.0784Output $0.2352
Flex
1 more label
86 tok/s94.88% uptime

Pricing

The average price customers actually pay for this model, next to the prices providers post. Caching and discounts mean the price actually paid is often well below the listed one.

Weighted average

Input

$0.01375/M tokens

Output

$0.6111/M tokens

DeepSeek V4.1 Flash 1017 — Price history

Price history

No effective price data for this selection

		$0.00446	$0.700	95.0%	3.2%
		$0.00759	$0.600	96.9%	16.4%
		$0.00846	$0.324	95.6%	14.1%
		$0.00933	$0.800	94.8%	1.7%
		$0.00934	$0.600	93.5%	0.7%
		$0.00994	$0.420	92.4%	2.8%
		$0.01035	$0.500	93.7%	4.0%
		$0.01072	$0.500	93.2%	2.6%
		$0.01086	$0.479	92.8%	1.2%
		$0.01128	$0.107	93.5%	1.2%
		$0.01159	$0.108	93.0%	1.0%
		$0.01205	$0.500	92.7%	6.5%
	Priority	$0.01210	$0.441	94.6%	2.4%
	Fast	$0.01234	$0.582	92.6%	0.4%
		$0.01337	$0.500	92.9%	3.0%
		$0.01503	$0.604	90.2%	5.1%
		$0.01505	$0.604	90.2%	1.0%
		$0.01530	$0.779	94.0%	0.2%
		$0.01588	$0.840	94.3%	1.2%
		$0.01606	$0.719	94.2%	3.4%
		$0.01621	$0.418	86.1%	1.4%
		$0.01650	$0.549	96.4%	0.2%
		$0.01652	$0.766	92.8%	0.9%
		$0.01679	$0.555	86.9%	4.1%
		$0.01759	$0.500	84.8%	1.0%
		$0.01762	$0.443	85.8%	0.1%
		$0.01967	$1.199	95.3%	6.9%
		$0.02000	$1.200	90.2%	0.8%
		$0.02026	$0.679	91.8%	0.1%
		$0.02311	$1.000	92.6%	0.5%
		$0.02442	$0.661	87.0%	0.4%
		$0.02450	$1.199	94.2%	0.0%
		$0.02455	$1.200	93.7%	2.7%
		$0.02478	$1.200	93.9%	1.5%
		$0.02619	$0.719	87.2%	0.1%
		$0.02665	$1.099	87.7%	0.0%
		$0.02765	$0.620	91.4%	1.2%
		$0.02963	$0.959	89.4%	0.0%
		$0.03183	$1.200	91.2%	1.1%
		$0.03831	$1.199	89.0%	0.0%
		$0.04832	$1.799	91.1%	0.5%
		$0.04987	$1.200	85.1%	0.2%
		$0.05219	$1.200	84.6%	0.3%
		$0.05823	$1.200	82.2%	0.5%
	Fast	$0.10792	$2.399	84.0%	0.4%
		$0.12149	$1.200	60.7%	0.2%
Not used in Nitro routing:Why these endpoints are not used
	Flex	$0.00815	$0.249	87.3%	2.0%
	Flex	$0.00883	$0.232	91.3%	0.3%
	Flex	$0.00550	$0.237	96.0%	0.6%

Effective in $0.00446Effective out $0.700
Token share
3.2%
Effective in $0.00759Effective out $0.600
Token share
16.4%
Effective in $0.00846Effective out $0.324
Token share
14.1%
Effective in $0.00933Effective out $0.800
Token share
1.7%
Effective in $0.00934Effective out $0.600
Token share
0.7%
Effective in $0.00994Effective out $0.420
Token share
2.8%
Effective in $0.01035Effective out $0.500
Token share
4.0%
Effective in $0.01072Effective out $0.500
Token share
2.6%
Effective in $0.01086Effective out $0.479
Token share
1.2%
Effective in $0.01128Effective out $0.107
Token share
1.2%
Effective in $0.01159Effective out $0.108
Token share
1.0%
Effective in $0.01205Effective out $0.500
Token share
6.5%
Effective in $0.01210Effective out $0.441
PriorityToken share
2.4%
Effective in $0.01234Effective out $0.582
FastToken share
0.4%
Effective in $0.01337Effective out $0.500
Token share
3.0%
Effective in $0.01503Effective out $0.604
Token share
5.1%
Effective in $0.01505Effective out $0.604
Token share
1.0%
Effective in $0.01530Effective out $0.779
Token share
0.2%
Effective in $0.01588Effective out $0.840
Token share
1.2%
Effective in $0.01606Effective out $0.719
Token share
3.4%
Effective in $0.01621Effective out $0.418
Token share
1.4%
Effective in $0.01650Effective out $0.549
Token share
0.2%
Effective in $0.01652Effective out $0.766
Token share
0.9%
Effective in $0.01679Effective out $0.555
Token share
4.1%
Effective in $0.01759Effective out $0.500
Token share
1.0%
Effective in $0.01762Effective out $0.443
Token share
0.1%
Effective in $0.01967Effective out $1.199
Token share
6.9%
Effective in $0.02000Effective out $1.200
Token share
0.8%
Effective in $0.02026Effective out $0.679
Token share
0.1%
Effective in $0.02311Effective out $1.000
Token share
0.5%
Effective in $0.02442Effective out $0.661
Token share
0.4%
Effective in $0.02450Effective out $1.199
Token share
0.0%
Effective in $0.02455Effective out $1.200
Token share
2.7%
Effective in $0.02478Effective out $1.200
Token share
1.5%
Effective in $0.02619Effective out $0.719
Token share
0.1%
Effective in $0.02665Effective out $1.099
Token share
0.0%
Effective in $0.02765Effective out $0.620
Token share
1.2%
Effective in $0.02963Effective out $0.959
Token share
0.0%
Effective in $0.03183Effective out $1.200
Token share
1.1%
Effective in $0.03831Effective out $1.199
Token share
0.0%
Effective in $0.04832Effective out $1.799
Token share
0.5%
Effective in $0.04987Effective out $1.200
Token share
0.2%
Effective in $0.05219Effective out $1.200
Token share
0.3%
Effective in $0.05823Effective out $1.200
Token share
0.5%
Effective in $0.10792Effective out $2.399
FastToken share
0.4%
Effective in $0.12149Effective out $1.200
Token share
0.2%
Not used in Nitro routing:Why these endpoints are not used
Effective in $0.00815Effective out $0.249
FlexToken share
2.0%
Effective in $0.00883Effective out $0.232
FlexToken share
0.3%
Effective in $0.00550Effective out $0.237
FlexToken share
0.6%

Performance

Throughput is how fast the model writes (tokens per second — higher is better). Latency is total round-trip time (lower is better). TTFT is time-to-first-token — how long before you see anything appear (lower is better).

Throughput

463tok/s

P50, best across providers

Latency

0.21s

P50, best provider

Reasoning effortLatency / throughput

AutoExacto benchmarks

GPQA DiamondTAU-BenchTenstorrent92.8%--PrimaLabs91.7%--Baseten (Fast)90.6%--AtlasCloud90.4%77.8%CoreWeave91.5%76.3%Baseten89.8%76.5%

+37 more providers

Uptime

Uptime is the percentage of the past 3 days that at least one provider was responding to requests. Availability is the percentage of time that inference was successfully served. OpenRouter continuously monitors and uses the next-best provider when one returns an error.

Uptime (3d)

100.00%

Availability (3d)

99.89%

Availability over the last 3 days

Last 72 hours

Availability 99.89%

3 Days Ago2 Days AgoYesterdayNow

Availability over the last 24 hours

OpenRouter Availability

99.87%

Without Routing

96.21%

When an error occurs in an upstream provider, we can recover by routing to another healthy provider, if your request filters allow it. You can access per-provider uptime data programmatically through the Endpoints API. Learn more about our load balancing and provider routing options.

Benchmarks

Scores on standardized evaluations and this model’s own outputs on shared benchmark prompts. Higher percentages are better — and each rank shows where this model lands among the models evaluated on that source.

Benchmark score summary for DeepSeek: DeepSeek V4.1 Flash 1017 (Artificial Analysis, Design Arena and Vals AI)
Source	Benchmark	Score
Artificial Analysis	DeepSeek V4.1 Flash 1017 (High) Intelligence Index	28.6
Artificial Analysis	DeepSeek V4.1 Flash 1017 (High) Terminal-Bench 4.0	28.8%
Artificial Analysis	DeepSeek V4.1 Flash 1017 (High) AA-Omniscience Accuracy	38.9%
Artificial Analysis	DeepSeek V4.1 Flash 1017 (High) AA-Omniscience Non-Hallucination Rate	38.8%
Artificial Analysis	DeepSeek V4.1 Flash (Max) HLE	39.2%
Artificial Analysis	DeepSeek V4.1 Flash (Max) AA-LCR	84.0%
Artificial Analysis	DeepSeek V4.1 Flash (Max) GDPval-AA	55.0%
Artificial Analysis	DeepSeek V4.1 Flash (Max) CritPt	14.3%
Artificial Analysis	DeepSeek V4.1 Flash (Max) SciCode	51.9%
Artificial Analysis	DeepSeek V4.1 Flash (Max) Terminal-Bench 4.0	26.8%
Artificial Analysis	DeepSeek V4.1 Flash 1017 (Non-reasoning) Intelligence Index	18.5
Artificial Analysis	DeepSeek V4.1 Flash 1017 (Non-reasoning) HLE	11.5%
Artificial Analysis	DeepSeek V4.1 Flash 1017 (Non-reasoning) AA-LCR	57.3%
Artificial Analysis	DeepSeek V4.1 Flash 1017 (Non-reasoning) CritPt	0.3%
Artificial Analysis	DeepSeek V4.1 Flash 1017 (Non-reasoning) SciCode	38.5%
Artificial Analysis	DeepSeek V4.1 Flash 1017 (Non-reasoning) Terminal-Bench 4.0	9.1%
Artificial Analysis	DeepSeek V4.1 Flash 1017 (Non-reasoning) AA-Omniscience Accuracy	21.9%
Artificial Analysis	DeepSeek V4.1 Flash 1017 (Non-reasoning) AA-Omniscience Non-Hallucination Rate	74.7%
Design Arena	DeepSeek-V4.1-Flash (1017) Agents Arena Spreadsheet Elo	1263
Design Arena	DeepSeek-V4.1-Flash (1017) Models Arena 3D Elo	1341
Design Arena	DeepSeek-V4.1-Flash (1017) Models Arena Asciiart Elo	1231
Design Arena	DeepSeek-V4.1-Flash (1017) Models Arena Code Categories Elo	1286
Design Arena	DeepSeek-V4.1-Flash (1017) Models Arena Data Visualization Elo	1264
Design Arena	DeepSeek-V4.1-Flash (1017) Models Arena Game Development Elo	1250
Design Arena	DeepSeek-V4.1-Flash (1017) Models Arena SVG Elo	1246
Design Arena	DeepSeek-V4.1-Flash (1017) Models Arena UI Component Elo	1284
Design Arena	DeepSeek-V4.1-Flash (1017) Models Arena Website Elo	1295
Vals AI	DeepSeek V4.1 Flash 1017 Access to Justice Benchmark	73.1%
Vals AI	DeepSeek V4.1 Flash 1017 Code Migration v1.1	67.5%
Vals AI	DeepSeek V4.1 Flash 1017 CyberOffense	54.8%
Vals AI	DeepSeek V4.1 Flash 1017 EMB v1.1	69.2%
Vals AI	DeepSeek V4.1 Flash 1017 Finance Agent v2.1	62.9%
Vals AI	DeepSeek V4.1 Flash 1017 Vals Finance Index	60.9%
Vals AI	DeepSeek V4.1 Flash 1017 HLAB-Vals-1.1	14.2%
Vals AI	DeepSeek V4.1 Flash 1017 Vals Legal Index	71.9%
Vals AI	DeepSeek V4.1 Flash 1017 Legal Research Bench	45.2%
Vals AI	DeepSeek V4.1 Flash 1017 Tax Agent Bench	44.4%
Vals AI	DeepSeek V4.1 Flash 1017 Terminal-Bench 4.0	31.8%
Vals AI	DeepSeek V4.1 Flash 1017 Vals Index	61.3%
Vals AI	DeepSeek V4.1 Flash 1017 Vibe Code Bench v1.1	87.7%

Apps

Public apps that send the most traffic to this model. Good signal for what real production workloads look like — and a hint at which use cases this model is best suited for.

Hermes Agent

Hermes Agent is an open-source, self-improving AI agent by Nous Research that runs persistently with memory across sessions, and builds reusable skills from experience. It comes with 40+ built-in tools, including web search, browser automation, and vision, plus scheduled automations and subagents.

10.9Ttokens

Claude Code

Claude Code is Anthropic's agentic coding tool that reads your entire codebase, plans and executes changes across files, runs tests, and iterates on failures, all from natural language prompts.

8.12Ttokens

There are many coding agents, but this one is yours.

6.99Ttokens

Strix

Open-source AI hackers that find and validate vulnerabilities in your code, APIs, and infrastructure, and open fix PRs.

6.02Ttokens

DeepSeek Harness

new

4.95Ttokens

Rest of OpenRouter

1.84T

Top 5 Apps

363B

DeepSeek: DeepSeek Flash versions

Not currently available through EU in-region routing. In-region routing docs

More models from DeepSeek

DeepSeek V4.1 Flash 1017

Text1.0M context$0.038 / $0.50

DeepSeek V4.1 Flash 1017 (batch)

Text1.0M context$0.05 / $0.15

DeepSeek Pro Latest

This model always redirects to the latest model in the DeepSeek Pro family.

Text1.0M context

DeepSeek Flash Latest

This model always redirects to the latest model in the DeepSeek Flash family.

Text1.0M context

DeepSeek V4.1 Flash

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.

Text1.0M context$0.0507 / $0.2028

DeepSeek V4.1 Flash (batch)

Text1.0M context$0.112 / $0.336

DeepSeek V4 Flash Vision Exp

DeepSeek V4 Flash Vision Exp is an experimental vision-enabled version of DeepSeek V4 Flash 0731 from DeepSeek, adding image understanding while matching the base model on text capabilities including agents, reasoning, and world knowledge. It is a sparse mixture-of-experts model with 13B active parameters out of 284B total.

It is suited for document and chart understanding, visual question answering, and multimodal agent workflows that interleave text and images.

Text1.0M context$0.2156 / $0.6468

DeepSeek V4 Pro 0813

DeepSeek V4 Pro 0813 is a large-scale mixture-of-experts model from DeepSeek. This is the GA release of DeepSeek V4 Pro.

Text1.0M context$0.0999 / $8

DeepSeek V4 Flash Latest

This model always redirects to the latest model in the DeepSeek V4 Flash family.

Text1.0M context

DeepSeek V4 Flash 0731

DeepSeek V4 Flash 0731 is a sparse mixture-of-experts model from DeepSeek, with 13B active parameters out of 284B total. This re-post-trained revision is suited for coding, reasoning, and agent workflows. This is the GA release of DeepSeek V4 Flash.

Text1.0M context$0.0434 / $0.30

DeepSeek V4 Flash 0731 (batch)

Text1.0M context$0.048 / $0.144

DeepSeek V4 Pro 0423

DeepSeek V4 Pro is a large-scale Mixture-of-Experts model from DeepSeek with 1.6T total parameters and 49B activated parameters, supporting a 1M-token context window. It is designed for advanced reasoning, coding, and long-horizon agent workflows, with strong performance across knowledge, math, and software engineering benchmarks.

Built on the same architecture as DeepSeek V4 Flash, it introduces a hybrid attention system for efficient long-context processing. Reasoning efforts high and xhigh are supported; xhigh maps to max reasoning. It is well suited for complex workloads such as full-codebase analysis, multi-step automation, and large-scale information synthesis, where both capability and efficiency are critical

Text1.0M context$0.3226 / $0.6452

DeepSeek V4 Flash 0423

DeepSeek V4 Flash is an efficiency-optimized Mixture-of-Experts model from DeepSeek with 284B total parameters and 13B activated parameters, supporting a 1M-token context window. It is designed for fast inference and high-throughput workloads, while maintaining strong reasoning and coding performance.

The model includes hybrid attention for efficient long-context processing. Reasoning efforts high and xhigh are supported; xhigh maps to max reasoning. It is well suited for applications such as coding assistants, chat systems, and agent workflows where responsiveness and cost efficiency are important.

Text1.0M context$0.0211 / $0.45

DeepSeek V3.2 Speciale

DeepSeek-V3.2-Speciale is a high-compute variant of DeepSeek-V3.2 optimized for maximum reasoning and agentic performance. It builds on DeepSeek Sparse Attention (DSA) for efficient long-context processing, then scales post-training reinforcement learning to push capability beyond the base model. Reported evaluations place Speciale ahead of GPT-5 on difficult reasoning workloads, with proficiency comparable to Gemini-3.0-Pro, while retaining strong coding and tool-use reliability. Like V3.2, it benefits from a large-scale agentic task synthesis pipeline that improves compliance and generalization in interactive environments.

Text131K context

DeepSeek V3.2

DeepSeek-V3.2 is a large language model designed to harmonize high computational efficiency with strong reasoning and agentic tool-use performance. It introduces DeepSeek Sparse Attention (DSA), a fine-grained sparse attention mechanism that reduces training and inference cost while preserving quality in long-context scenarios. A scalable reinforcement learning post-training framework further improves reasoning, with reported performance in the GPT-5 class, and the model has demonstrated gold-medal results on the 2025 IMO and IOI. V3.2 also uses a large-scale agentic task synthesis pipeline to better integrate reasoning into tool-use settings, boosting compliance and generalization in interactive environments.

Users can control the reasoning behaviour with the reasoning enabled boolean. Learn more in our docs

Text164K context$0.2088 / $0.3096

DeepSeek V3.2 Exp

DeepSeek-V3.2-Exp is an experimental large language model released by DeepSeek as an intermediate step between V3.1 and future architectures. It introduces DeepSeek Sparse Attention (DSA), a fine-grained sparse attention mechanism designed to improve training and inference efficiency in long-context scenarios while maintaining output quality. Users can control the reasoning behaviour with the reasoning enabled boolean. Learn more in our docs

The model was trained under conditions aligned with V3.1-Terminus to enable direct comparison. Benchmarking shows performance roughly on par with V3.1 across reasoning, coding, and agentic tool-use tasks, with minor tradeoffs and gains depending on the domain. This release focuses on validating architectural optimizations for extended context lengths rather than advancing raw task accuracy, making it primarily a research-oriented model for exploring efficient transformer designs.

Text164K context$0.27 / $0.41

DeepSeek V3.1 Terminus

DeepSeek-V3.1 Terminus is an update to DeepSeek V3.1 that maintains the model's original capabilities while addressing issues reported by users, including language consistency and agent capabilities, further optimizing the model's performance in coding and search agents. It is a large hybrid reasoning model (671B parameters, 37B active) that supports both thinking and non-thinking modes. It extends the DeepSeek-V3 base with a two-phase long-context training process, reaching up to 128K tokens, and uses FP8 microscaling for efficient inference. Users can control the reasoning behaviour with the reasoning enabled boolean. Learn more in our docs

The model improves tool use, code generation, and reasoning efficiency, achieving performance comparable to DeepSeek-R1 on difficult benchmarks while responding more quickly. It supports structured tool calling, code agents, and search agents, making it suitable for research, coding, and agentic workflows.

Text164K context$0.27 / $1

DeepSeek V3.1

DeepSeek-V3.1 is a large hybrid reasoning model (671B parameters, 37B active) that supports both thinking and non-thinking modes via prompt templates. It extends the DeepSeek-V3 base with a two-phase long-context training process, reaching up to 128K tokens, and uses FP8 microscaling for efficient inference. Users can control the reasoning behaviour with the reasoning enabled boolean. Learn more in our docs

It succeeds the DeepSeek V3-0324 model and performs well on a variety of tasks.

Text164K context$0.25 / $0.95

DeepSeek V3.1 Base

This is a base model, trained only for raw next-token prediction. Unlike instruct/chat models, it has not been fine-tuned to follow user instructions. Prompts need to be written more like training text or examples rather than simple requests (e.g., “Translate the following sentence…” instead of just “Translate this”).

DeepSeek-V3.1 Base is a 671B parameter open Mixture-of-Experts (MoE) language model with 37B active parameters per forward pass and a context length of 128K tokens. Trained on 14.8T tokens using FP8 mixed precision, it achieves high training efficiency and stability, with strong performance across language, reasoning, math, and coding tasks.

Text164K context

R1 Distill Qwen 7B

DeepSeek-R1-Distill-Qwen-7B is a 7 billion parameter dense language model distilled from DeepSeek-R1, leveraging reinforcement learning-enhanced reasoning data generated by DeepSeek's larger models. The distillation process transfers advanced reasoning, math, and code capabilities into a smaller, more efficient model architecture based on Qwen2.5-Math-7B. This model demonstrates strong performance across mathematical benchmarks (92.8% pass@1 on MATH-500), coding tasks (Codeforces rating 1189), and general reasoning (49.1% pass@1 on GPQA Diamond), achieving competitive accuracy relative to larger models while maintaining smaller inference costs.

Text131K context

DeepSeek: DeepSeek V4.1 Flash 1017 (nitro)

Providers