KymaKyma
DeepSeek

DeepSeek

DeepSeek V3

#4 on Kyma

DeepSeek's previous-generation flagship, still carrying heavy production traffic on Kyma — the stable, proven choice for frontier-open quality at balanced pricing.

Modalities

Text → Text

Below release

fp4

Weights

Published

deepseek-ai/DeepSeek-V3.2

Where it sits in the catalogue Kyma measures

Every number against every text model Kyma prices per token — a stated rule, not a chosen line-up.

Price · input#19/70

$0.351

output $0.513

median $0.977 · best $0.0494

Price · output#10/70

$0.513

per 1M tokens generated

median $3.531 · best $0.2164

Availability · 30d#63/68

83.09%

544 observations since 2026-05-02

median 96.2% · best 100.0%

Throughput#67/68

11.5 tok/s

probe median, one fixed prompt

median 40.8 tok/s · best 308.3 tok/s

Response time#10/68

1.20 s

probe median, to a complete answer

median 2.50 s · best 0.52 s

Context

160K

max output 8,192

published by DeepSeek, not measured here

Tick above each rail is this model, below it the other 69. Dashed rule is the field median, solid is its best. Better is left; the axis stops at the 90th percentile, so a few models sit past its right edge.

Usage

How much this model is actually called here.

Rank

#4

of 113 active models

Tokens served

181.0M

all-time

Platform share

5.6%

of all tokens

Tokens · last 15 daysAug 21 → Sep 4

How it behaves under real clients

The same model answering different prompt shapes — measured, not benchmarked.

ClientRequestsTokensThroughputTo first tokenp95 totalCompleted
OpenAI SDK368368.0K7 tok/s5.61 s32 s23.6%

Grouped by the client that sent the request. Not a benchmark: the same model on different prompt shapes. Throughput and time-to-first-token are medians; p95 is the slowest response in twenty.

Pricing

Pay per token. Cached input bills at this model’s own cached rate, listed below.

$0.351 /1M input$0.513 /1M output
Hobby10 req/day · 2K in / 500 out
~$0.29/mo
Production1,000 req/day · 2K in / 500 out
~$28.75/mo
Scale20,000 req/day · 2K in / 500 out
~$575/mo
+ Estimate your workload
1,000
2,000
500

Estimated monthly cost

$28.75

$0.9585 / day on DeepSeek V3

Same workload on:

DeepSeek V4 Flash Vision$31.18+8%
DeepSeek V4 Pro$62.11+116%

Estimates use list pricing. Actual bills depend on real token counts, and every response includes its exact cost.

When to use DeepSeek V3

Updated 2026-06-10

Where this model earns its cost — and where it doesn't.

DeepSeek V3 is the previous-generation flagship from DeepSeek, built for reasoning, coding, and general-purpose work. It sits in Kyma's frontier-open quality tier, and its appeal is exactly what the label says: stable and proven — a fit for teams that value predictable behavior over chasing the latest checkpoint.

Coding agents lead its traffic, followed by direct Python and Node.js integrations. Every request gets the platform's automatic failover, and each response reports its exact cost in usage.cost.

The 160K-token context window pairs with function calling, structured outputs, and extended reasoning, so it can run multi-step agent loops, not just single-turn chat. It sits in the frontier-open tier as the value pick — see the pricing table above for the current rate.

Coding agents

Its top production app on Kyma by tokens is OpenClaw, a coding agent — code generation, review, and refactoring inside agent harnesses.

Reasoning pipelines

Extended reasoning support handles analysis and planning tasks that need more than pattern matching.

Cost-conscious production

Frontier-open quality at balanced pricing makes it a sensible default for high-volume general workloads where value matters.

Structured extraction

Structured outputs plus the 160K context let it pull reliable JSON from long documents and codebases in one pass.

Tool-calling backends

Function calling support makes it a dependable engine behind Python and Node.js services that orchestrate tools.

Not ideal for: Image inputs (it's text-only), latency-critical UX where its medium speed tier shows, or single responses longer than the 8K output cap.

How it compares

Against the peers people actually weigh it against.

SpecDeepSeek V3DeepSeek V4 Flash VisionDeepSeek V4 Pro
Input /1M$0.351$0.297$0.6901
Output /1M$0.513$0.891$1.38
Context160K1M1M
ToolsYesYesYes
ReasoningYesYesYes
Throughput11.5 tok/s148.7 tok/s31.7 tok/s

Quick start

Up and running in under two minutes.

  1. 1

    Create an API key

    Sign up and grab a key from the dashboard — $0.50 free credit, no card required.

    Get API key →
  2. 2

    Make your first request

    Drop in your key and send a chat completion — fully OpenAI-compatible.

    curl https://kymaapi.com/v1/chat/completions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-v3",
    "messages": [
      {"role": "user", "content": "Explain prompt caching in one paragraph."}
    ]
  }'
  3. 3

    Stream responses

    Add "stream": true to receive tokens as they arrive.

    curl https://kymaapi.com/v1/chat/completions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-v3",
    "stream": true,
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

FAQ

Common questions about this model.

What is the context window of DeepSeek V3?

DeepSeek V3 has a 160K-token context window — roughly 235 pages of text in a single request.

How much does the DeepSeek V3 API cost?

$0.351 per 1M input tokens and $0.513 per 1M output tokens. No subscription; you pay only for what you use.

Does DeepSeek V3 support function calling?

Yes — DeepSeek V3 supports tool/function calling and structured outputs (JSON mode), so it works with agent frameworks out of the box.

Are the weights for DeepSeek V3 publicly available?

Yes. DeepSeek publishes DeepSeek V3's weights as deepseek-ai/DeepSeek-V3.2, so you can download and run the model yourself (https://huggingface.co/deepseek-ai/DeepSeek-V3.2, read 2026-08-14). Kyma serves it because it is convenient and has failover behind it, not because it is the only way to reach it.

Is DeepSeek V3 ever served below the precision its creator released it at?

Sometimes. DeepSeek V3 was released by DeepSeek at fp8 + bf16 (https://huggingface.co/deepseek-ai/DeepSeek-V3.2, read 2026-08-14), and at least one route serving it here reports fp4 — narrower than that. Weights compressed below the release usually answer close to it, but it is not the identical artefact, and nothing in the response tells you which one answered. Kyma checked the routes on 2026-09-03, so it is your call rather than a silent one.

How do I use DeepSeek V3?

Kyma is OpenAI-compatible: point your SDK's base URL at https://kymaapi.com/v1, use your Kyma API key, and set the model to deepseek-v3. Signing up is free and includes $0.50 of credit — no card required.

Is DeepSeek V3 good for coding?

Yes — coding is one of its recommended workloads, and its top app on Kyma by tokens is OpenClaw, a coding agent. Tool calling plus structured outputs keep agent loops on track. The 8K output cap means very long single-file generations are better split into steps.

Why pick a previous-generation flagship over something newer?

Stability. DeepSeek V3 has served enough production traffic on Kyma for its behavior to be well understood. If your pipeline depends on consistent outputs more than peak capability, the proven option is often the right one — and it comes at balanced pricing.

Why use DeepSeek V3 through Kyma?

One API key covers this and every other model on the platform through an OpenAI-compatible endpoint. Every request gets automatic failover if a serving path degrades, every response includes its exact cost in usage.cost, and signup comes with $0.50 of free credit — no card required.

Start with $0.50 free credit — no card required.Create account →

More models by DeepSeek

ModelContextInputOutput
DeepSeekDeepSeek V4 Flash Vision1M$0.297$0.891
DeepSeekDeepSeek V4 Flash1M$0.1389$0.2778
DeepSeekDeepSeek V4 Pro1M$0.6901$1.38
DeepSeekDeepSeek R164K$0.7425$2.957