deepseek

DeepSeek Coder V2 236B

Released 2024-06-17Context 163.84Kdeepseek-coder-v2-236b

PipelineScore

85.0MAINLINE

Ranked #5 of 52 models · 90th percentileAn even spread: no standout, no liability (84 to 89 across all five categories). Best-fit profile: Local-first.

Category breakdown

Score per category, normalized 0–100 against the v1 anchor.

Code

87.2

Reason

83.7

Tool Use

84.8

RAG

89.0

Speed

88.7

Strengths

RAG89.0

Speed88.7

Code87.2

Same model, different rigs

Every submission of DeepSeek Coder V2 236B on the 0–100 scale. The spread is the point: where it runs changes what you get.

0255075100

Best 87.3 on lab-h200-141gb · lowest 84.5 on 2x-rtx-4090 · spread 2.8 pts across 4 runs. Hover a dot for its rig.

Sample tasks

A taste of what the test pack measures. Full prompts are private and rotated daily.

CodeDifficulty 1code-fib-1

Fibonacci function

Write a Python `fib(n)` returning the nth Fibonacci number, O(n).

ReasonDifficulty 1reason-math-1

Train meeting time

Two trains, opposite directions, given speeds and start times — when do they meet?

RAGDifficulty 2rag-extract-1

Extract metrics to JSON

From the context, extract net sales, operating margin, and free cash flow as a JSON object. Numbers only.

Tool UseDifficulty 2tool-schema-1

OpenAPI param selection

Given an OpenAPI schema with limit/offset/sort, fill JSON for 'next 50, recent first.'

RAGDifficulty 2rag-grounding-1

Refuses to fabricate

Context lacks the answer — does the model fabricate or correctly say it can't?

Compare with

DeepSeek Coder V2 236B vs openai/gpt-oss-20b DeepSeek Coder V2 236B vs DeepSeek R1 671B-A37B DeepSeek Coder V2 236B vs Qwen 3 235B-A22B MoE DeepSeek Coder V2 236B vs DeepSeek V3 671B-A37B DeepSeek Coder V2 236B vs Qwen 2.5 72B Instruct