DeepSeek
Deepseek V3 2 ReasoningOpen
Deepseek V3 2 Reasoning by DeepSeek.
Intelligence
38.0 #463
Top 60% tier
Output Speed
40 t/s
ITL 25.0 ms/tok
Blended price (3:1)
$0.478 /1M
In $0.27/M · Out $1.1/M
Context
131K
≈ 328 A4 pages · max out 16,384
Quality Domain Radar
6-Axis Frontier Evaluation vs Category Average & Flagship
Deepseek V3 2 Reasoning
Claude Mythos Preview (#1)
Average (68)
Radar axes are approximated from category scores where direct benchmark measurements are missing — see Methodology.
Strengths & Trade-offs
Where it wins
- Open weights — self-host, fine-tune and audit with no vendor lock-in
- Very low blended price — $0.478/1M mixed tokens
- Cheap prompt-cache reads at $0.027/1M
Trade-offs to watch
- Legacy tier — a newer successor is available
Killer Use Cases
Whole-repo analysis
Single-pass ingestion of large codebases, contracts or research corpora.
SWE-bench agents
Issue-to-PR resolution across multi-file repos — refactor, test and verify autonomously.
Quantitative pipelines
Numeric reasoning for finance, science and engineering calculations.
Developer Quickstart & API
Set
DEEPSEEK_API_KEY first — get an API keycurl https://api.deepseek.com/v1/chat/completions \
-H "Authorization: Bearer $DEEPSEEK_API_KEY" \
-H "content-type: application/json" \
-d '{
"model": "deepseek-v3-2-reasoning",
"messages": [{"role": "user", "content": "Refactor this function for O(n) time and explain the change."}]
}'import os
from openai import OpenAI
client = OpenAI(
base_url="https://api.deepseek.com/v1",
api_key=os.environ["DEEPSEEK_API_KEY"],
)
resp = client.chat.completions.create(
model="deepseek-v3-2-reasoning",
messages=[{"role": "user", "content": "Refactor this function for O(n) time and explain the change."}],
)
print(resp.choices[0].message.content)const res = await fetch("https://api.deepseek.com/v1/chat/completions", {
method: "POST",
headers: {
Authorization: `Bearer ${process.env.DEEPSEEK_API_KEY}`,
"content-type": "application/json",
},
body: JSON.stringify({
model: "deepseek-v3-2-reasoning",
messages: [{ role: "user", content: "Refactor this function for O(n) time and explain the change." }],
}),
});
console.log((await res.json()).choices[0].message.content);Official Resources & Links
API deployment
Compare hosts & latency
Throughput, TTFT and cache pricing across Groq, Cerebras, Together, DeepInfra, Fireworks, Bedrock and Azure.
GroqCerebrasTogether AIDeepInfraFireworks AIAWS BedrockAzure AI
Scores by Source
Category Scores
Benchmark Results
Live Specs & Architecture
Under the hood
Mixture-of-Experts · 671B (37B active MoE)
- Mixture-of-Experts routing — only a fraction of total parameters is active per token.
- Prompt caching supported — cached input reads at $0.027/1M tokens.
Release date
Jan 1, 2025
Architecture
Mixture-of-Experts · 671B (37B active MoE)
Parameters
Not disclosed
License
Open Weights
Context Window
131,000 tokens
Max Output
16,384 tokens
Modalities
text
Cache Pricing
read $0.027 · write $0.3375 /1M
Reasoning
Effort levels
Input Price
$0.27/M
Output Price
$1.1/M
Output Speed
40 t/s
Time to First Token
—
Data Updated
Aug 14, 2026
Estimated monthly bill
Estimated monthly spend
$0.00
Calculating savings vs GPT-4o...
Cite & embed
GitHub Markdown badge
Loading badge markdown...BibTeX
@misc{llmpodium2026deepseekv32reasoning,
title = {Benchmark Analysis and Intelligence Evaluation of Deepseek V3 2 Reasoning},
author = {LLMPodium Research},
year = {2026},
url = {https://llmpodium.com/models/deepseek-v3-2-reasoning}
}Embed Score Badge
Add a live Podium Score badge to your GitHub or Hugging Face README — it links back to the live ranking.
Preview
Markdown
Loading badge snippet...HTML
Loading badge snippet...More from DeepSeek
Alternative Models
Closest alternatives to Deepseek V3 2 Reasoning based on capabilities, pricing, and benchmark scores.