whatmodel.sh / DeepSeek-Coder-V2 Instruct (0614)

DeepSeek-Coder-V2 Instruct (0614)

+ contribute
Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

DeepSeek-Coder-V2 Instruct (0614) is an open-weights language model from DeepSeek, released June 2024. Its published size is 236B parameters (21B active per token). Licence: deepseek.

Catalogue identity notes

Comparison-only. Arena entry 'deepseek-coder-v2' (text). LMArena's leaderboard table (arena-leaderboard space @6f2eeeb) labels 'deepseek-coder-v2' as 'DeepSeek-Coder-V2-Instruct' (DeepSeek AI, DeepSeek License), linking https://huggingface.co/deepseek-ai/DeepSeek-Coder-V2-Instruct. Maker repo deepseek-ai/DeepSeek-Coder-V2-Instruct @2453c79: licence tag other (deepseek-license). Parameters: DeepSeek-V2 architecture: 236B total / 21B active (HF safetensors metadata 235.7B). Released 2024-06: month the maker's Hugging Face repo was created (createdAt 2024-06-14). Original June 2024 release (API name DeepSeek-Coder-V2-0614); LMArena links deepseek-ai/DeepSeek-Coder-V2-Instruct. The 0724 update is a separate repo (DeepSeek-Coder-V2-Instruct-0724).

53.0±0.7#298 of 719 LLMsfrom 12 benchmark results across 8 sources

$ how it compares

The LLMs ranked either side of DeepSeek-Coder-V2 Instruct (0614), out of 828 with a score, then more from DeepSeek.

#modelscorewhereRAMreleased
356Qwen Turbo (2024-11-01) Alibaba53±2comparison only2024-11
357Qwen2.5-VL 32B Instruct Alibaba53±2local~22GB2025-03
358DeepSeek-Coder-V2 Instruct (0614) DeepSeek53±1local~92GB2024-06
359s1.1 32B Simple Scaling53±1local~22GB2025-02
360DeepSeek-Coder-V2 Instruct (0724) DeepSeek53±1local~151GB2024-09
more from DeepSeek
21DeepSeek V4.1 Flash DeepSeek78±1api $0.26/M +local~537GB2026-09
31DeepSeek V4 Pro (0813) DeepSeek76±1api $1.63/M +local~1010GB2026-08
35DeepSeek V3.2 Speciale DeepSeek76±3comparison only

see all LLMs · find the ones that fit your machine

$ what it's good at

coding▁▁▁▃▆██▇▆▄▃▁ #212 of 458 · 55% of its score · 6 tests

Beats Qwen3.5 4B, Qwen3-VL 32B Instruct, Solar Open 100B (102B-A12B) here, though they rank higher overall. Trails QwQ-32B-Preview, Ministral 3 3B Reasoning 2512, DeepSeek-Coder-V2 Instruct (0724) here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
HumanEval85.4% ✎▁·▁▁▃▄▆▅█▃▁▁ beats 91/99easy · avg 79%▇▇▇▇▇▇▇▇ 20%sharp
HumanEval+82.3% ✎▁·▁▁▁▃▄▄▄█▃▁ beats 94/99easy · avg 74%▇▇▇▇▇▇▇ 18%sharp
LiveCodeBench official (Jan 1–Nov 1, 2024)34.4% ✎▁▁·▁▂▄▆█▅▂·▂ beats 19/31hard · avg 26%▇▇▇▇ 10%sharp · at its level
MBPP+75.1% ✎▁···▁▁▂▅▇█▆▃ beats 72/75easy · avg 68%▇▇ 4.0%at its level
WebApp1K-React pass@1 (onekq-ai)70.0% ✎▃▁▁█▁▄▅▃▄▇▁▄ beats 18/31medium · avg 44%▇ 1.9%sharp · at its level · noisy
SciCode main problems (official leaderboard)3.1% ✎▃···█··▃·▃·▂ beats 11/17very hard · avg 1.1%▇ 1.5%sharp · too hard for it
chat & writing▁▁▁▃▅▄▅█▆▆▇▃ #259 of 393 · 18% of its score · 3 tests

Beats Ministral 3 14B Reasoning 2512, Ministral 3 14B, OpenThinker2 32B here, though they rank higher overall. Trails Flan-T5 XXL, Granite 4.0 H 1B, Granite 4.0 1B here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
Arena-Hard (v0.1, GPT-4-1106 judge)62.3% ✎▂▁▁▅█▅▄▃▃▅▃▁ beats 42/60medium · avg 52%▇▇▇▇ 10%sharp · at its level
Polyrating rating (bias-aware re-fit of Chatbot Arena votes, mid-2024)1181.0 ✎▂▁▂▁▃▇█▆▇▄▂▂ beats 76/95avg 1201▇▇ 4.0%blunt
LMArena (text)1265.3 ✎▁▁▁▂▃▃▃▆▆▇█▃ beats 100/379avg 1268▇▇ 3.9%blunt
math▁▂▃▅▆▅▄█▄▁▁▁ #262 of 398 * · 3.4% of its score · 1 test

Beats Molmo 2 8B, Llama 3.3 70B Instruct, DeepSeek-V2.5 here, though they rank higher overall. Trails DeepSeek R1 Distill Qwen 1.5B, Mixtral 8x7B v0.1 (base), InternVL2-4B here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
GSM8K (ZeroEval, zero-shot)93.8% ✎▁▄·▁▄█▅▅▅███ beats 30/43very easy · avg 89%▇ 3.4%sharp · maxed out
other tests2 not in a task · 23% of its score
testresultvs the others in its rundifficultycountswhy
MMLU-Redux (ZeroEval, 2,778 questions)79.6% ✎▁·▁▁▄▄▄▅█▄▂▁ beats 26/53easy · avg 77%▇▇▇▇▇▇ 15%typical
ZebraLogic (ZeroEval, 1,000 puzzles)21.1% ✎▁▁▃▆██▁▁·▁▁▁ beats 32/57hard · avg 18%▇▇▇ 7.6%sharp

$ where to run it

llama.cpp · recommended

version

  1. Install llama.cpp · Homebrew and winget installs name the commands llama-cli and llama-server instead.
    macOS / Linux
    curl -LsSf https://llama.app/install.sh | sh
    Windows
    irm https://llama.app/install.ps1 | iex

    other ways to install ↗

  2. Run DeepSeek-Coder-V2 Instruct (0614) · needs ~142.45 GB · from huggingface.co/bartowski
    chat in the terminal
    llama cli -hf bartowski/DeepSeek-Coder-V2-Instruct-GGUF:Q4_K_M
  3. Or start a local server and open the address it prints (http://localhost:9931)
    llama serve -hf bartowski/DeepSeek-Coder-V2-Instruct-GGUF:Q4_K_M

LM Studio

version

  1. Install LM Studio · Open the app once before using the lms command.

    download it (macOS / Windows / Linux) ↗

  2. Run DeepSeek-Coder-V2 Instruct (0614) · needs ~142.45 GB · from huggingface.co/bartowski

    open in LM Studio ↗

  3. Or download it in the terminal
    lms get https://huggingface.co/bartowski/DeepSeek-Coder-V2-Instruct-GGUF@Q4_K_M -y
  4. Then run lms ls to see its name, and lms chat followed by that name

MLX (Mac)

  1. Install MLX (Mac) · Apple silicon Macs only, text only.
    with uv
    uv tool install mlx-lm
    or
    pip install mlx-lm
  2. Run DeepSeek-Coder-V2 Instruct (0614) · needs ~132.7 GB, Mac only · from huggingface.co/GGorman
    chat in the terminal
    mlx_lm.chat --model GGorman/DeepSeek-Coder-V2-Instruct-Q4-mlx
  3. Or start an OpenAI-compatible server at http://localhost:8080/v1
    mlx_lm.server --model GGorman/DeepSeek-Coder-V2-Instruct-Q4-mlx

Weights and documentation: huggingface.co/deepseek-ai ↗

$ how the score was worked out · EvalPlus / EvalPerf published… 42%, WildEval ZeroEval published r… 26%, Arena-Hard-Auto v0.1 leaderbo… 10% and 5 more

Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works

sourceresultsshare of the scorelicence
EvalPlus / EvalPerf published evaluation runs3▇▇▇▇▇▇▇▇▇▇▇▇▇ 42%Apache-2.0
WildEval ZeroEval published results3▇▇▇▇▇▇▇▇ 26%Apache-2.0
Arena-Hard-Auto v0.1 leaderboard (repository README, pre-v2)1▇▇▇ 10%Apache-2.0
LiveCodeBench official website results (2024-12-26)1▇▇▇ 10%CC-BY-SA-4.0
Polyrating: A Cost-Effective and Bias-Aware Rating System for LLM Evaluation (arXiv:2409.00696v3)1▇ 4%CC-BY-4.0
LMArena leaderboard dataset1▇ 4%CC BY 4.0
WebApp1K models leaderboard (onekq-ai)1▇ 2%MIT
SciCode leaderboard (repository README)1▇ 2%Apache-2.0
prior · pull toward the middle▇ 0%
weighted average= 53.0

$ improve this page

Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.

data updated · sources and licences · json