Tech Meridian← ALL MODELS
PROMY MERIDIAN

DEEPSEEK · MODEL RELEASE TRACKER

DeepSeek-V3.1-Base

DeepSeek-V3.1-Base is a released DeepSeek model (announced 2025-08-21). It introduces hybrid inference modes (Think / Non-Think), improved agent/tool-use and multi-step reasoning, continued pretraining on 840B tokens to extend long-context capability, an updated tokenizer/chat template, support for the Anthropic API format and Beta Strict Function Calling, open-source weights on Hugging Face, availability via chat.deepseek.com, and upcoming pricing changes starting Sep 5, 2025.

CURRENT SNAPSHOT2/5 DIMENSIONS WITH DATA

The dimensions that change the decision.

PRICING

Not established from the available sources.

CONTEXT WINDOW

Not established from the available sources.

MODALITIES

Not established from the available sources.

BENCHMARKS
Performance vs previous modelDeepSeek-V3.1-Think reaches answers in less time versus DeepSeek-R1-0528 (reported faster thinking).
AVAILABILITY
AvailabilityAvailable now via chat.deepseek.com (toggle Think/Non-Think with the 'DeepThink' button).
Open-source weightsOpen-source weights available on Hugging Face: DeepSeek-V3.1-Base and DeepSeek-V3.1.

RELEASE TIMELINE

Published, source-backed release events only.

1 SOURCES · IMPORTANCE 7.0

DeepSeek releases DeepSeek‑V3.1 with hybrid 'Think & Non‑Think' agent mode and open-source weights

DeepSeek released DeepSeek‑V3.1, introducing a hybrid inference mode called Think & Non‑Think (toggleable via the “DeepThink” button at chat.deepseek.com), faster 'Think' responses vs. DeepSeek‑R1‑0528, improved tool use and multi-step agent skills, Anthropic API format support, and beta Strict Function Calling support. V3.1 Base received 840B tokens of continued pretraining for long-context extension; the tokenizer and chat template were updated (tokenizer_config.json); open-source weights for DeepSeek‑V3.1 and DeepSeek‑V3.1‑Base are published on Hugging Face, and new pricing/off-peak discount changes take effect Sep 5, 2025, 16:00 UTC.

→

VERIFIABLE FACTS

Every value stays attached to a source and date.

BENCHMARKS · Performance vs previous modelDEVELOPER CLAIM

DeepSeek-V3.1-Think reaches answers in less time versus DeepSeek-R1-0528 (reported faster thinking).

CAPABILITIES · Agent and reasoning capabilitiesDEVELOPER CLAIM

Post-training boosts tool use and multi-step agent tasks; stronger multi-step reasoning for complex search tasks.

CAPABILITIES · Tokenizer and chat templateDEVELOPER CLAIM

Tokenizer and chat template updated — new tokenizer config: tokenizer_config.json.

WHAT CHANGED

Stored passport versions, without reconstructed history.

Passport created8 facts