DeepSeek Coder V2 Lite

Name: DeepSeek Coder V2 Lite
Author: DeepSeek

Newer version available: DeepSeek V4 Pro

Released

2024-06-17

Last refreshed

2026-10-17

Status

Researched 10d ago

Open weightsCommercial use: permittedCodingLong contextClassification

DeepSeek releases · 10 in the last 12 monthsChangelog →

Specifications

Family: DeepSeek Coder V2
Released: 2024-06-17
Context: 128K
Parameters: 16B
Architecture: Mixture of Experts
Knowledge cutoff: 2023-11
Specialization: code
Openness: Open weights
License: DeepSeek LicensePermitted
Weights: Available
Code: Unknown
Training: Fine-tuned

Created by

DeepSeek

Advancing artificial general intelligence (AGI).

Hangzhou, Zhejiang, China

Founded 2023

Website

Pricing

Output / 1M

$0.500

Input / 1M

$0.500

Cheapest of 1 route · Fireworks AI

Providers(1)

Fireworks AI

About

DeepSeek Coder V2 Lite is an open-source Mixture-of-Experts (MoE) language model specifically tailored for efficiency and cost-effectiveness in coding tasks. It operates with a 15.7B parameter count, but only 2.4B are active at any given time, making it comparable to GPT4-Turbo for code-centric applications. This model supports 338 programming languages and has an extended context length of 128K tokens, facilitating the handling of complex codebases and lengthy prompts. Its features encompass code generation, completion, understanding, and mathematical reasoning, making it versatile for diverse coding applications.

Top use-case fit: coding, agents, and build tasks

Coding

Q/$ B

Classification

Q/$ C

Providers(1)

Compare API pricing across 1 tracked provider route for input and output tokens, batch, and cached reads when available.

Provider	Input / 1M	Output / 1M	Route
Fireworks AI	$0.500	$0.500	ServerlessProvisioned

Available via routers & gateways(1)

OpenRouter

Hybrid

Unified hybrid gateway to 400+ models from 60+ providers via a single OpenAI-compatible API, with optional auto-routing that selects the best model per prompt.

PassthroughFireworks AI

Capabilities

No model capability flags are currently sourced.

Benchmark peer barsfor Coding

HumanEvalRank 38 of 99

Claude Sonnet 4.6

98.0

96.7

Claude Opus 4.6

95.0

Grok-3

94.5

DeepSeek Coder V2 Litecurrent

81.1

Benchmark scores(4)

Scores are benchmark-specific and are direction-aware: the same numeric gap can mean very different outcomes across suites. Use the leaderboard context and this model's provider route to decide whether the winning margin is meaningful for your workload.

Benchmark	Score	Version	Evaluation	Source
Google-Proof Q&A	52.4	diamondObserved 2026-03-06	—	Source
HellaSwag	91.4	10-shotObserved 2026-03-06	—	Source
HumanEval	81.1	pass@1Observed 2026-03-06	—	Source
Massive Multitask Language Understanding	78.9	5-shotObserved 2026-03-06	—	Source