DeepSeek-V3

deepseek-ai/DeepSeek-V3

CHAT High tier Chat Open-weight

DeepSeek-V3 (deepseek-ai/DeepSeek-V3) is a chat model from together with a 65.5K-token context window; open-weight; capabilities tools; input text; output text.

65.5KContext

8.2KMax output

$1.25$ / 1M in

14.2Intelligence

Pricing

Input: $1.25 / 1M tokens
Output: $1.25 / 1M tokens
Currency: USD
Note: Indicative US list price — verify with the vendor.

Indicative US list price — a reference only, verify with the vendor. Source: litellm. Last verified 2026-07-22.

Benchmarks

Intelligence index: 14.2
Coding: 23
Math: 26

Cited from a public third-party leaderboard — a reference, verify at the source. Source: Artificial Analysis. Last verified 2026-07-24.

Specifications

Context window: 65,536 tokens (65.5K)
Max output: 8,192 tokens
Weights: Open-weight
Input modalities: text
Output modalities: text
Release date: 2024-12-26

Capabilities

tools

Provenance

committedlitellmmodels-devoverrides

Last verified: 2026-10-30

Related models

Qwen3 235B A22B Instruct

Qwen/Qwen3-235B-A22B-Instruct-2507-tput

together · CHAT · 262K ctx

Qwen3 Coder 480B

Qwen/Qwen3-Coder-480B-A35B-Instruct-FP8

together · CHAT · 256K ctx

Llama 3.3 70B Instruct Turbo

meta-llama/Llama-3.3-70B-Instruct-Turbo

together · CHAT · 131.1K ctx

DeepSeek-R1

deepseek-ai/DeepSeek-R1

together · CHAT · 128K ctx

Llama 4 Maverick 17B

meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8

together · CHAT

Markdown · Vendor JSON · Full catalog · All together models