Databricks
Frontier

DBRX Instruct

Open

Databricks' open-weights MoE, 132B total / 36B active. Tuned for instruction following on Databricks Mosaic AI.

DBRX Instruct is a frontier AI model from Databricks. It costs $0.7500 per million input tokens and $2.25 per million output tokens (blended $1.20/M), with a 32,000-token context window.

RatesUSD per million tokens. Blend is 70% input, 30% output, our calculation.

INPUT
$0.7500/M
per million input tokens
OUTPUT
$2.25/M
per million output tokens
BLENDED 70/30
$1.20/M
unchanged since 3 May
CONTEXT
32,000
tokens

Hand-maintained from Databricks's published pricing; no check date on record.

What it is good at
  • Open-weights MoE
  • Tight Databricks integration
  • 32K context
Typical use cases
  • Databricks-native chat
  • Self-hosted MoE research

Benchmarks

vs. best public score
MMLU73.7%
Multitask academic knowledge across 57 subjects. Source: Databricks (self-reported), MMLU 5-shot, DBRX Instruct, read 2026-11-10.
HumanEval70.1%
Python function synthesis from docstrings. Source: Databricks (self-reported), HumanEval (0-shot pass@1, temp 0), DBRX Instruct, read 2026-11-10.
LMArena Elo1195 Elo
Crowd-sourced head-to-head preference Elo rating. Source: LMArena (independent), LMArena Text Arena score, read 2026-11-10.
Hand-curated from each provider's published reports and public leaderboards. Methodology varies across sources, treat as directional rather than authoritative.

More from Databricks

See all 3 →

Frequently asked questions

How much does DBRX Instruct cost?

DBRX Instruct costs $0.7500 per million input tokens and $2.25 per million output tokens, for a blended reference rate of $1.20 per million tokens.

What is DBRX Instruct's context window?

DBRX Instruct supports up to 32,000 tokens of context in a single request.

What is DBRX Instruct best for?

DBRX Instruct is well suited to Open-weights MoE, Tight Databricks integration and 32K context.

Who makes DBRX Instruct?

DBRX Instruct is developed and served by Databricks. It was released in Mar 2024.

Terms used on this page