Model Tracker

DeepSeek-V4.1-Flash

DeepSeek-V4.1-Flash is an AI model made by DeepSeek AI. DeepSeek AI released it on 10 September 2026. It is generally available.

In inglese

DeepSeek AI

DeepSeek-V4.1-Flash is an AI model made by DeepSeek AI. DeepSeek AI released it on 10 September 2026. It is generally available. It takes text, image and returns text. Through the API it costs $0.30 per 1M tokens for input and $1.20 per 1M tokens for output. Its weights are open. Its licence is MIT.

Released: 10 September 2026 — api-docs.deepseek.com, read 16 October 2026
Status: Generally available — api-docs.deepseek.com, read 16 October 2026
Takes: Text, Image — api-docs.deepseek.com, read 16 October 2026
Returns: Text — api-docs.deepseek.com, read 16 October 2026
Open weights: Yes — huggingface.co, read 16 October 2026
Licence: MIT — huggingface.co, read 16 October 2026
Context window: Not confirmed: two readings disagree: 1048576 and 1000000
Longest answer: Not confirmed: two readings disagree: 393216 and 384000
Knowledge cutoff: Not confirmed: The maker's reviewed model card, release notes, and API documentation do not state a knowledge cutoff for this model.
API price, input: $0.30 per 1M tokens — api-docs.deepseek.com, read 16 October 2026
API price, output: $1.20 per 1M tokens — api-docs.deepseek.com, read 16 October 2026
Where: API, Downloadable weights — api-docs.deepseek.com, read 16 October 2026
Parameters: Not confirmed: The model card's technical report gives 552B backbone parameters, while the same page's model metadata says 763B params. The maker does not reconcile these figures or specify which is the model's parameter count for this field.
Model card: https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash — huggingface.co, read 16 October 2026

Benchmarks

Measured by others, not by the maker: each score with the name the measurer gives the model and the setting it ran, its day and its licence.

Arena

Text Arena, deepseek-v4.1-flash-max: 1,475 (1,468–1,481, 12,793 votes), published 15 October 2026
Vision Arena, deepseek-v4.1-flash-max: 1,260 (1,249–1,271, 3,559 votes), published 9 October 2026

Arena’s leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0: ratings from people’s votes between two anonymous models, as Arena publishes them. Data · CC BY 4.0

Epoch AI

Epoch Capabilities Index, DeepSeek V4.1 Flash: 155.6 (153.5–157.6), as of 19 October 2026
GPQA Diamond, deepseek-v4.1-flash_max: 89.8% (± 1.7), run 9 October 2026
FrontierMath, tiers 1–3, deepseek-v4.1-flash_max: 67.4% (± 2.8), run 8 October 2026
FrontierMath, tier 4, deepseek-v4.1-flash_max: 26.8% (± 7.0), run 8 October 2026
OTIS Mock AIME 2024–2025, deepseek-v4.1-flash_max: 98.3% (± 0.6), run 8 October 2026
SimpleQA Verified, deepseek-v4.1-flash_max: 42.9% (± 1.6), run 13 October 2026

Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks, CC BY 4.0: Epoch AI’s own runs and its Capabilities Index. Data · CC BY 4.0

Questions

When was DeepSeek-V4.1-Flash released?

DeepSeek AI released DeepSeek-V4.1-Flash on 10 September 2026.

Is DeepSeek-V4.1-Flash still available?

Yes: it is generally available, as DeepSeek AI states it.

How much does the DeepSeek-V4.1-Flash API cost?

On the standard tier, DeepSeek-V4.1-Flash costs $0.30 per 1M tokens for input and $1.20 per 1M tokens for output, as DeepSeek AI lists it.

Are the weights of DeepSeek-V4.1-Flash open?

Yes: DeepSeek AI publishes the weights of DeepSeek-V4.1-Flash for download.

What is the licence of DeepSeek-V4.1-Flash?

The licence of DeepSeek-V4.1-Flash is MIT, as DeepSeek AI names it.

What does DeepSeek-V4.1-Flash take and return?

DeepSeek-V4.1-Flash takes text, image and returns text.

Sources

Change Log (read 16 October 2026)
Lists Models (read 16 October 2026)
Models & Pricing (read 16 October 2026)
DeepSeek-V4.1-Flash (read 16 October 2026)

All of the Model Tracker