DeepSeek · Open weights
DeepSeek Latest
Compatibility alias that routes to the newest thinking DeepSeek model. Currently routes to DeepSeek V4 Pro 0813 Thinking, the general-availability release built for coding, tool use, cybersecurity, automation, and long-horizon agent workflows. Requests are routed through privacy-friendly third-party providers and are not sent directly to DeepSeek.
- Reasoning
- Tools
- Structured output
- Input
- Text
- Output
- Text
- Reasoning effort
- none · low · high · max
- Discovery tier
- Open weights
Pricing
USD per 1M tokens- Input
- $0.23
- Output
- $0.69
- Cache read
- $0.0073
Reference prices as published by the model provider. API example
Specifications
- Context window
- 1M
- Max output
- 384K
Side by side
DeepSeek Latest compared
Specifications, prices, speed and benchmark scores next to each other. Benchmarks below show every compared model too.
| DeepSeek LatestDeepSeekThis model | DeepSeek V3.2 ThinkingDeepSeekRemove | |
|---|---|---|
| Provider | DeepSeek | DeepSeek |
| Discovery tier | Open weights | Open weights |
| Added | May 3, 2026 | Dec 1, 2025 |
| Limits | ||
| Context window | 1.05MBest | 163K |
| Max output | 384KBest | 65.54K |
| Average output | — | 1.4K tokens |
| Pricing, USD per 1M tokens | ||
| Input | $0.23 | $0.21Best |
| Output | $0.69 | $0.31Best |
| Cache read | $0.0073Best | $0.0216 |
| Performance | ||
| Throughput | 74.3 tok/sBest | 41.1 tok/s |
| Time to first token | 2.7 s | 1.7 sBest |
| Request success (24h) | 99.7% | 99.9%Best |
| Benchmark indexes | ||
| Intelligence Index | 36.0Best | 21.5 |
| Coding Index | 68.8Best | 44.2 |
| Math Index | — | 92.0 |
| Capabilities | ||
| Reasoning | Yes | Yes |
| Tools | Yes | Yes |
| Structured output | Yes | Yes |
| PDF input | No | Yes |
| Reasoning effort | none, low, high, max | — |
Public evaluations
Benchmarks
Independent evaluation results, with how each score ranks among the models in the Noumen catalog.
- DeepSeek Latest
- DeepSeek V3.2 Thinking
Intelligence Index
- DeepSeek Latest36.0
- DeepSeek V3.2 Thinking21.5
DeepSeek Latest tested at max reasoning effort
DeepSeek Latest: Better than 79% of 326 scored models
Indexes
Coding Index
Composite of the coding evaluations
- DeepSeek Latest: 68.8Top
- DeepSeek V3.2 Thinking: 44.2
Math Index
Composite of the math evaluations
- DeepSeek Latest: —
- DeepSeek V3.2 Thinking: 92.0
Agentic work
AutomationBench-AA
Workflow automation, penalized for guardrail violations
- DeepSeek Latest: 56.7%
- DeepSeek V3.2 Thinking: —
AutomationBench-AA tasks completed
Workflows finished end to end without a guardrail violation
- DeepSeek Latest: 25.9%
- DeepSeek V3.2 Thinking: —
AA-Briefcasev1.1
Agentic knowledge work, rated head to head
- DeepSeek Latest: 1,256 Elo
- DeepSeek V3.2 Thinking: —
GDPval-AAv2.1
Economically valuable occupational tasks, rated head to head
- DeepSeek Latest: 1,449 EloTop
- DeepSeek V3.2 Thinking: 691 Elo
Document reasoning
GDP.pdf
Professional reasoning over PDF documents, all-pass rate
- DeepSeek Latest: 11.4%
- DeepSeek V3.2 Thinking: —
AA-LCRv1.1
Reasoning across long documents
- DeepSeek Latest: 80.3%Top
- DeepSeek V3.2 Thinking: 73.3%
Reasoning and knowledge
Humanity’s Last Exam
Expert-level questions across academic fields
- DeepSeek Latest: 41.0%Top
- DeepSeek V3.2 Thinking: 24.6%
GPQA Diamond
Graduate-level science questions
- DeepSeek Latest: 92.8%Top
- DeepSeek V3.2 Thinking: 84.0%
MMLU-Pro
Multi-subject knowledge and reasoning
- DeepSeek Latest: —
- DeepSeek V3.2 Thinking: 86.2%
IFBench
Following precise output instructions
- DeepSeek Latest: —
- DeepSeek V3.2 Thinking: 60.7%
Coding
Terminal-Benchv4.0
Practical software engineering in a terminal
- DeepSeek Latest: 14.1%
- DeepSeek V3.2 Thinking: —
SciCode
Scientific computing in Python
- DeepSeek Latest: 51.0%
- DeepSeek V3.2 Thinking: —
LiveCodeBench
Recent competitive programming problems
- DeepSeek Latest: —
- DeepSeek V3.2 Thinking: 86.2%
Math
AIME 2025
Competition mathematics, 2025 problems
- DeepSeek Latest: —
- DeepSeek V3.2 Thinking: 92.0%
Legacy evaluations
Terminal-Bench Hard
Earlier terminal task suite
- DeepSeek Latest: —
- DeepSeek V3.2 Thinking: 35.6%
τ²-Bench Telecom
Tool use in a customer-support simulation
- DeepSeek Latest: —
- DeepSeek V3.2 Thinking: 90.6%
AA-LCR (unversioned)
Long-context reasoning, earlier grading
- DeepSeek Latest: 80.3%Top
- DeepSeek V3.2 Thinking: 73.3%
Scores from Artificial Analysis, updated Oct 26, 2026. DeepSeek Latest was tested at max reasoning effort; compare scores measured at the same effort.Artificial Analysis
Live routing data
Performance and reliability
Recent averages across requests routed for this model. Your own speed depends on prompt length, reasoning effort and load.
- Throughput
- 74.3 tok/s
- Faster than 50% of 250 models
- Time to first token
- 2.7 s
- Quicker than 62% of 250 models
- Prompt cache hits
- 71.3%
- About $0.0711 per 1M input tokens with caching
- Request success (24h)
- 99.7%
- Reliable
Cache hits are the share of input tokens recently served from a prompt cache across all requests. They are not a forecast of your own hit rate.
- Healthy (98% or more)
- Degraded (90–98%)
- Down (under 90%)
- No requests
Developers
Use it through the API
Noumen’s API is OpenAI-compatible: any Chat Completions client works with a base URL and a key.
curl https://www.noumen.xyz/api/v1/chat/completions \
-H "Authorization: Bearer $NOUMEN_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "deepseek/deepseek-latest",
"reasoning_effort": "none",
"messages": [{ "role": "user", "content": "Explain this stack trace." }]
}'import os
from openai import OpenAI
client = OpenAI(base_url="https://www.noumen.xyz/api/v1", api_key=os.environ["NOUMEN_API_KEY"])
reply = client.chat.completions.create(
model="deepseek/deepseek-latest",
messages=[{"role": "user", "content": "Explain this stack trace."}],
)
print(reply.choices[0].message.content)DeepSeek
Related models
Other DeepSeek models, starting with the same family.