# DeepSeek-Coder-V2-Lite-Instruct by DeepSeek: Open Model Source: https://savrn.com/models/deepseek-coder-v2-lite-instruct Markdown alternate of the page above; the site index is https://savrn.com/llms.txt --- ## Runs On What it takes to serve DeepSeek-Coder-V2-Lite-Instruct (15.7B parameters): the memory its weights need at each precision, and the cheapest way to rent enough data-center GPUs to hold them. | Precision | Weights | Memory needed | Cheapest setup | Per hour | Also fits | | --- | --- | --- | --- | --- | --- | | 16-bit | 31.4 GB | 37.7 GB | 1x [MI300X](https://savrn.com/ai-index/pricing/gpus/mi300x) (192 GB) Vultr | $1.85 | [1x H100](https://savrn.com/ai-index/pricing/gpus/h100) $1.99 · [1x MI325X](https://savrn.com/ai-index/pricing/gpus/mi325x) $2.00 | | 8-bit | 15.7 GB | 18.8 GB | 1x [MI300X](https://savrn.com/ai-index/pricing/gpus/mi300x) (192 GB) Vultr | $1.85 | [1x H100](https://savrn.com/ai-index/pricing/gpus/h100) $1.99 · [1x MI325X](https://savrn.com/ai-index/pricing/gpus/mi325x) $2.00 | | 4-bit | 7.9 GB | 9.4 GB | 1x [MI300X](https://savrn.com/ai-index/pricing/gpus/mi300x) (192 GB) Vultr | $1.85 | [1x H100](https://savrn.com/ai-index/pricing/gpus/h100) $1.99 · [1x MI325X](https://savrn.com/ai-index/pricing/gpus/mi325x) $2.00 | Memory is the weights at that precision plus 20% for the runtime and a short context; a long context needs more. Prices are the lowest on-demand hourly rates in the [SAVRN Index](https://savrn.com/ai-index/pricing/gpus), read Oct 25, 2026. [DeepSeek-Coder-V2-Lite-Instruct on every accelerator the SAVRN Index prices, at every precision](https://savrn.com/models/deepseek-coder-v2-lite-instruct/gpus) ## SAVRN's Notes on DeepSeek-Coder-V2-Lite-Instruct Sixty-four routed experts, six active per token, is the line to read first in this coding model, because only six fire per token yet all 15.7 billion parameters must sit in memory: 37.7 GB at 16-bit for 31.4 GB of safetensors, 18.8 GB at 8-bit, 9.4 GB at 4-bit. On the cheapest Index setup, one MI300X with 192 GB at $1.85 per hour, even the full-precision copy uses a fifth of the card. DeepSeek built it on DeepseekV2ForCausalLM with a 163,840-token context. The license field reads other and the file carries no summary, so there is no commercial-use answer to hand you; read the publisher's license text before any deployment. Also check that the only evaluation on file is a third-party MMLU-Pro result of 41.57 from EvalEval, that our Index lists no host prices per million tokens, and that the paper is arXiv:2401.06066 with a last update of 2024-07-03. SAVRN Research, 2026-09-18 ## Model Card We present DeepSeek-Coder-V2, an open-source Mixture-of-Experts (MoE) code language model that achieves performance comparable to GPT4-Turbo in code-specific tasks. Specifically, DeepSeek-Coder-V2 is further pre-trained from an intermediate checkpoint of DeepSeek-V2 with additional 6 trillion tokens. Through this continued pre-training, DeepSeek-Coder-V2 substantially enhances the coding and mathematical reasoning capabilities of DeepSeek-V2, while maintaining comparable performance in general language tasks. Compared to DeepSeek-Coder-33B, DeepSeek-Coder-V2 demonstrates significant advancements in various aspects of code-related tasks, as well as reasoning and general capabilities.… Excerpt from the card by DeepSeek, licensed other. ## Configuration Architecture DeepseekV2ForCausalLM Context length (tokens) 163,840 Layers 27 Hidden size 2,048 Feed-forward size 10,944 Attention heads 16 Key/value heads 16 Vocabulary size 102,400 Routed experts 64 Experts active per token 6 RoPE base 10,000 Stored precision bfloat16 Model type deepseek_v2 ## Identity and Version Repository deepseek-ai/DeepSeek-Coder-V2-Lite-Instruct Publisher DeepSeek Task Text generation Modality Text Library transformers Parameters 15.7B parameters Languages Not stated by the source Revision e434a23f91ba5b4923cf6c9d9a238eb4a08e3a11 First published 2024-06-14 Last updated 2024-07-03 ## Files and Weights 14 files, 31.4 GB in total. The weights are 4 files totalling 31.4 GB in safetensors. Weights4 files · 31.4 GB Configuration6 files · 572.0 KB Tokenizer2 files · 4.6 MB Documentation1 file · 11.3 KB Repository1 file · 1.5 KB Every file | File | Type | Size | SHA-256 | | --- | --- | --- | --- | | model-00001-of-000004.safetensors | Weights | 8.6 GB | 75d08ddaf92b | | model-00002-of-000004.safetensors | Weights | 8.6 GB | 7bf22dfa2715 | | model-00003-of-000004.safetensors | Weights | 8.6 GB | 18f5a20f4d73 | | model-00004-of-000004.safetensors | Weights | 5.6 GB | 1365ca25494e | | config.json | Configuration | 1.5 KB | — | | configuration_deepseek.py | Configuration | 10.3 KB | — | | generation_config.json | Configuration | 181 B | — | | model.safetensors.index.json | Configuration | 479.9 KB | — | | modeling_deepseek.py | Configuration | 78.7 KB | — | | tokenization_deepseek_fast.py | Configuration | 1.4 KB | — | | README.md | Documentation | 11.3 KB | — | | .gitattributes | Repository | 1.5 KB | — | | tokenizer.json | Tokenizer | 4.6 MB | — | | tokenizer_config.json | Tokenizer | 1.3 KB | — | ## License and Download License other Access Open weights, no gate Download size 31.4 GB [Download from DeepSeek](https://huggingface.co/deepseek-ai/DeepSeek-Coder-V2-Lite-Instruct) Released by DeepSeek through its official repository on Hugging Face. ## Built From - Described by [arXiv:2401.06066](https://savrn.com/papers/deepseekmoe-towards-ultimate-expert-specialization-in-mixture-of-experts-language-models) ## Evaluations Each result is shown as reported, with the conditions its reporter stated. None is a SAVRN measurement. A comparison lines two results up only when their configuration, unit and setup are all stated and identical. | Benchmark | Conditions | Result | Reported by | Revision | Date | | --- | --- | --- | --- | --- | --- | | TIGER-Lab/MMLU-Pro | Task mmlu_proMetric mmlu_proComparison conditions not established | 41.57 | [EvalEval](https://huggingface.co/datasets/evaleval/EEE_datastore/blob/192329fb7d6b15b7b0936a1a58ae862aa7e8ba24/flat/objects/0c/0f/0c0f446a-6c2a-47c7-a120-4fafb90f92a3.json) Reported by a third party | Evaluated revision not stated | 2026-06-30 | ## Memory Requirements | Precision | Weights in memory | | --- | --- | | As published | 31.4 GB | | 16-bit | 31.4 GB | | 8-bit | 15.7 GB | | 4-bit | 7.9 GB | Weights only, from the published parameter count; the key-value cache and runtime add to this. ## Questions About DeepSeek-Coder-V2-Lite-Instruct ### How much GPU memory does DeepSeek-Coder-V2-Lite-Instruct need? About 37.7 GB at 16-bit and 9.4 GB at 4-bit: the weights (15.7B parameters) plus a working margin. A long context needs more. ### What is the cheapest GPU to run DeepSeek-Coder-V2-Lite-Instruct on? At 16-bit, 1x MI300X from $1.85 an hour; at 4-bit, 1x MI300X from $1.85 an hour, at the lowest on-demand prices the SAVRN Index lists. ### What license is DeepSeek-Coder-V2-Lite-Instruct released under? other, as its publisher declares it. Read the license text before commercial use. ### What is DeepSeek-Coder-V2-Lite-Instruct's context length? 163,840 tokens, from the maximum position embeddings in its published configuration. ## Similar Models Model · Text generation ### [Keural-MoE-14B](https://savrn.com/models/keural-moe-14b) [MKD Co., Ltd.](https://savrn.com/model-publishers/mkd-dev) Open weights apache-2.0 15.4B parameters transformers [View model](https://savrn.com/models/keural-moe-14b) Model · Text generation ### [Qwen2.5-Coder-14B-Instruct](https://savrn.com/models/qwen2-5-coder-14b-instruct) [Qwen](https://savrn.com/model-publishers/qwen) Qwen2.5-Coder is the latest series of Code-Specific Qwen large language models (formerly known as CodeQwen). As of now, Qwen2.5-Coder has covered six mainstream model sizes, 0.5, 1.5, 3, 7, 14, 32 billion parameters, to meet the needs of different developers. Qwen2.5-Coder brings the following improvements upon CodeQwen1.5: - Significantly improvements in code generation, code reasoning and code fixing. Base on the strong Qwen2.5, we scale up the training tokens into 5.5 trillion including source code, text-code grounding, Synthetic data, etc. Qwen2.5-Coder-32B has become the current state-of-the-art open-source codeLLM, with its coding abilities matching those of GPT-4o. - A more… Open weights apache-2.0 14.8B parameters 32,768 tokens transformers [View model](https://savrn.com/models/qwen2-5-coder-14b-instruct) Model · Text generation ### [Qwen2.5-14B-Instruct-AWQ](https://savrn.com/models/qwen2-5-14b-instruct-awq) [Qwen](https://savrn.com/model-publishers/qwen) Qwen2.5 is the latest series of Qwen large language models. For Qwen2.5, we release a number of base language models and instruction-tuned language models ranging from 0.5 to 72 billion parameters. Qwen2.5 brings the following improvements upon Qwen2: - Significantly more knowledge and has greatly improved capabilities in coding and mathematics, thanks to our specialized expert models in these domains. - Significant improvements in instruction following, generating long texts (over 8K tokens), understanding structured data (e.g, tables), and generating structured outputs especially JSON. More resilient to the diversity of system prompts, enhancing role-play implementation and… Open weights apache-2.0 14.8B parameters 32,768 tokens transformers [View model](https://savrn.com/models/qwen2-5-14b-instruct-awq) Model · Text generation ### [Qwen2.5-Coder-14B-Instruct-AWQ](https://savrn.com/models/qwen2-5-coder-14b-instruct-awq) [Qwen](https://savrn.com/model-publishers/qwen) Qwen2.5-Coder is the latest series of Code-Specific Qwen large language models (formerly known as CodeQwen). As of now, Qwen2.5-Coder has covered six mainstream model sizes, 0.5, 1.5, 3, 7, 14, 32 billion parameters, to meet the needs of different developers. Qwen2.5-Coder brings the following improvements upon CodeQwen1.5: - Significantly improvements in code generation, code reasoning and code fixing. Base on the strong Qwen2.5, we scale up the training tokens into 5.5 trillion including source code, text-code grounding, Synthetic data, etc. Qwen2.5-Coder-32B has become the current state-of-the-art open-source codeLLM, with its coding abilities matching those of GPT-4o. - A more… Open weights apache-2.0 14.8B parameters 32,768 tokens transformers [View model](https://savrn.com/models/qwen2-5-coder-14b-instruct-awq) Model · Text generation ### [Qwen2.5-14B-Instruct](https://savrn.com/models/qwen2-5-14b-instruct) [Qwen](https://savrn.com/model-publishers/qwen) Qwen2.5 is the latest series of Qwen large language models. For Qwen2.5, we release a number of base language models and instruction-tuned language models ranging from 0.5 to 72 billion parameters. Qwen2.5 brings the following improvements upon Qwen2: - Significantly more knowledge and has greatly improved capabilities in coding and mathematics, thanks to our specialized expert models in these domains. - Significant improvements in instruction following, generating long texts (over 8K tokens), understanding structured data (e.g, tables), and generating structured outputs especially JSON. More resilient to the diversity of system prompts, enhancing role-play implementation and… Open weights apache-2.0 14.8B parameters 32,768 tokens transformers [View model](https://savrn.com/models/qwen2-5-14b-instruct) Model · Text generation ### [JiRackUltra_14b](https://savrn.com/models/jirackultra-14b) [Center Business Solutions inc](https://savrn.com/model-publishers/cmsmanhattan) A fast and efficient 14B model optimized for CPU inference. The model was refactored with BitNet features and an updated tokenizer that includes new Routing, Media, Vision, Sound, Tool call, and Robotics tags. Built on a DeepSeek R1-14B architecture with native ternary (BitNet-style) support and ready-to-run GGUF quantizations. - JiRack is a cloud-ready model that helps save money on cloud infrastructure. It can be used as an expert model in RAG deployments, with the ONNX JiRack Java server as an alternative. - Benefits high quality CPU inference TQ2 on Llama.cpp and Ollama via QAT - Robotcs, Routing, Coding, Multimedia, Advanced tool calling via CMSManhattan/JiRackPrecisionTokenizer… Open weights mit 14.8B parameters 131,072 tokens [View model](https://savrn.com/models/jirackultra-14b) ## DeepSeek [All models and datasets](https://savrn.com/model-publishers/deepseek-ai) ## Versions - [e434a23f91ba](https://savrn.com/models/deepseek-coder-v2-lite-instruct/versions/e434a23f91ba) · current 2026-09-18 ## Explore More - [All text generation models](https://savrn.com/models/tasks/text-generation) - [Model comparisons](https://savrn.com/models/comparisons) - [The model directory](https://savrn.com/models) - [Open model prices by host](https://savrn.com/ai-index/pricing/open-models) ## Source - Repository metadata, read 2026-10-19. - [DeepSeek website](https://www.deepseek.com/) - [Hugging Face record](https://huggingface.co/deepseek-ai/DeepSeek-Coder-V2-Lite-Instruct) - [How the hub is built](https://savrn.com/model-hub/methodology)