# Qwen3-8B by Qwen: Open-Weight Model
Source: https://savrn.com/models/qwen3-8b
Markdown alternate of the page above; the site index is https://savrn.com/llms.txt

---

## Runs On

What it takes to serve Qwen3-8B (8.2B 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 | 16.4 GB | 19.7 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 | 8.2 GB | 9.8 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 | 4.1 GB | 4.9 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 22, 2026.

[Qwen3-8B on every accelerator the SAVRN Index prices, at every precision](https://savrn.com/models/qwen3-8b/gpus)

## SAVRN's Notes on Qwen3-8B

At 16-bit the weights are 16.4 GB and the model needs 19.7 GB, which fits one MI300X with 192 GB, the Index's cheapest setup at $1.85 an hour on-demand. At 8-bit the need drops to 9.8 GB and at 4-bit to 4.9 GB, so the question is never which card but how many copies to stack on one. With 8.2 billion parameters, a 40,960-token context and a switch between thinking and non-thinking modes, this is a text model for everyday work.

The license is the easy part: Apache 2.0 permits commercial use, modification and redistribution, with notices and any NOTICE file kept and significant changes stated, plus an express patent grant. Two things to weigh: it derives from Qwen3-8B-Base, so decide whether you want this checkpoint or the base, and compare the hourly card against the one Index host price, Nscale at $0.07 in and $0.18 out per million tokens.

SAVRN Research, 2026-09-18

## Model Card

By Qwen, published under apache-2.0, revision b968826d9c46.

### Qwen3 Highlights

Qwen3 is the latest generation of large language models in Qwen series, offering a comprehensive suite of dense and mixture-of-experts (MoE) models. Built upon extensive training, Qwen3 delivers groundbreaking advancements in reasoning, instruction-following, agent capabilities, and multilingual support, with the following key features:

[Read the full model card (1,884 words)](https://savrn.com/models/qwen3-8b/card)

## Configuration

Architecture

Qwen3ForCausalLM

Context length (tokens)

40,960

Layers

36

Hidden size

4,096

Feed-forward size

12,288

Attention heads

32

Key/value heads

8

Head dimension

128

Vocabulary size

151,936

RoPE base

1,000,000

Stored precision

bfloat16

Model type

qwen3

## Identity and Version

Repository

Qwen/Qwen3-8B

Publisher

Qwen

Task

Text generation

Modality

Text

Library

transformers

Parameters

8.2B parameters

Languages

Not stated by the source

Revision

b968826d9c46dd6066d109eabc6255188de91218

First published

2025-04-27

Last updated

2025-07-26

## Files and Weights

15 files, 16.4 GB in total. The weights are 5 files totalling 16.4 GB in safetensors.

Weights5 files · 16.4 GB

Configuration3 files · 33.8 KB

Tokenizer4 files · 15.9 MB

Documentation2 files · 28.0 KB

Repository1 file · 1.6 KB

Every file

| File | Type | Size | SHA-256 |
| --- | --- | --- | --- |
| model-00001-of-00005.safetensors | Weights | 4.0 GB | 31d6a825ae35 |
| model-00002-of-00005.safetensors | Weights | 4.0 GB | 5991236cea6f |
| model-00003-of-00005.safetensors | Weights | 4.0 GB | c5185c4794be |
| model-00004-of-00005.safetensors | Weights | 3.2 GB | b5ee7de71fbf |
| model-00005-of-00005.safetensors | Weights | 1.2 GB | 20c2d6366ab8 |
| config.json | Configuration | 728 B | — |
| generation_config.json | Configuration | 239 B | — |
| model.safetensors.index.json | Configuration | 32.9 KB | — |
| LICENSE | Documentation | 11.3 KB | — |
| README.md | Documentation | 16.7 KB | — |
| .gitattributes | Repository | 1.6 KB | — |
| merges.txt | Tokenizer | 1.7 MB | — |
| tokenizer.json | Tokenizer | 11.4 MB | aeb13307a71a |
| tokenizer_config.json | Tokenizer | 9.7 KB | — |
| vocab.json | Tokenizer | 2.8 MB | — |

## License and Download

License

apache-2.0

Access

Open weights, no gate

Download size

16.4 GB

[Download from Qwen](https://modelscope.cn/models/Qwen/Qwen3-8B)

Released by Qwen through ModelScope. [Read the license](https://www.apache.org/licenses/LICENSE-2.0).

## Built From

- Derived from Qwen/Qwen3-8B-Base
- Described by [arXiv:2309.00071](https://savrn.com/papers/yarn-efficient-context-window-extension-of-large-language-models)
- Described by [arXiv:2505.09388](https://savrn.com/papers/qwen3-technical-report)

## 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 |
| --- | --- | --- | --- | --- | --- |
| LiquidAI/ifstruct-v1.0 | Task ifstruct_v1Metric ifstruct_v1Comparison conditions not established | 79.75 | Liquid AI — IFStruct v1.0 blog (Qwen3-8B) Reported by a third party | Evaluated revision not stated | 2026-06-30 |

## Memory Requirements

| Precision | Weights in memory |
| --- | --- |
| As published | 16.4 GB |
| 16-bit | 16.4 GB |
| 8-bit | 8.2 GB |
| 4-bit | 4.1 GB |

Weights only, from the published parameter count; the key-value cache and runtime add to this.

## Hosted Prices

| Host | Input / output | Unit | Observed |
| --- | --- | --- | --- |
| Nscale | $0.07 / $0.18 | input / output, per million tokens | Oct 22, 2026 |

From the [SAVRN Index](https://savrn.com/ai-index/pricing/open-models).

## Built on This Model

- Quantized from[Qwen3-8B-AWQ](https://savrn.com/models/qwen3-8b-awq)
- Derived from[Qwen3-8B-AWQ](https://savrn.com/models/qwen3-8b-awq)
- Quantized from[Qwen3-8B-FP8](https://savrn.com/models/qwen3-8b-fp8)
- Derived from[Qwen3-8B-FP8](https://savrn.com/models/qwen3-8b-fp8)
- Derived from[Symbiotic-8B](https://savrn.com/models/symbiotic-8b)
- Derived from[ProactiveInquirer-Qwen3-8B-Merged](https://savrn.com/models/proactiveinquirer-qwen3-8b-merged)
- Adapter of[ProactiveInquirer-Qwen3-8B](https://savrn.com/models/proactiveinquirer-qwen3-8b)
- Derived from[ProactiveInquirer-Qwen3-8B](https://savrn.com/models/proactiveinquirer-qwen3-8b)
- Adapter of[sat-tutor-qwen3-8b](https://savrn.com/models/sat-tutor-qwen3-8b)
- Derived from[sat-tutor-qwen3-8b](https://savrn.com/models/sat-tutor-qwen3-8b)
- Adapter of[qwen3-8b-followup-lora](https://savrn.com/models/qwen3-8b-followup-lora)
- Derived from[qwen3-8b-followup-lora](https://savrn.com/models/qwen3-8b-followup-lora)
- Derived from[CHILLGuard-8B](https://savrn.com/models/chillguard-8b)
- Derived from[qwen3-8b-sft-agentclinic-diagnosis-shaped-masked](https://savrn.com/models/qwen3-8b-sft-agentclinic-diagnosis-shaped-masked)
- Derived from[mira-agent-instinct](https://savrn.com/models/mira-agent-instinct)
- Adapter of[qwen-Manimator-1-grpo](https://savrn.com/models/qwen-manimator-1-grpo)
- Derived from[qwen-Manimator-1-grpo](https://savrn.com/models/qwen-manimator-1-grpo)
- Derived from[ALFWorld-Qwen3-8B-Adaptive-Random-Each](https://savrn.com/models/alfworld-qwen3-8b-adaptive-random-each)
- Quantized from[sat-tutor-qwen3-8b-gguf](https://savrn.com/models/sat-tutor-qwen3-8b-gguf)
- Derived from[sat-tutor-qwen3-8b-gguf](https://savrn.com/models/sat-tutor-qwen3-8b-gguf)
- Derived from[agentevolver-appworld-qwen3-8b-base-cot](https://savrn.com/models/agentevolver-appworld-qwen3-8b-base-cot)
- Derived from[qwen3-8b-sft-261001-bigmath-sol-fsdp-0-epoch-4](https://savrn.com/models/qwen3-8b-sft-261001-bigmath-sol-fsdp-0-epoch-4)
- Adapter of[qwen3-8b-worker-farca](https://savrn.com/models/qwen3-8b-worker-farca)
- Derived from[qwen3-8b-worker-farca](https://savrn.com/models/qwen3-8b-worker-farca)
- Adapter of[RubriX-8B](https://savrn.com/models/rubrix-8b)
- Derived from[RubriX-8B](https://savrn.com/models/rubrix-8b)
- Adapter of[qwen3-8b-moe-merge-merged](https://savrn.com/models/qwen3-8b-moe-merge-merged)
- Derived from[qwen3-8b-moe-merge-merged](https://savrn.com/models/qwen3-8b-moe-merge-merged)
- Adapter of[lora-factory](https://savrn.com/models/lora-factory)
- Derived from[lora-factory](https://savrn.com/models/lora-factory)
- Derived from[dense-1-8](https://savrn.com/models/dense-1-8)

## Compare Qwen3-8B

- [DeepSeek-R1-0528-Qwen3-8B vs Qwen3-8B](https://savrn.com/models/compare/deepseek-r1-0528-qwen3-8b-vs-qwen3-8b)
- [Llama-3.1-8B-Instruct vs Qwen3-8B](https://savrn.com/models/compare/llama-3-1-8b-instruct-vs-qwen3-8b)
- [Meta-Llama-3-8B-Instruct vs Qwen3-8B](https://savrn.com/models/compare/meta-llama-3-8b-instruct-vs-qwen3-8b)

## Questions About Qwen3-8B

### How much GPU memory does Qwen3-8B need?

About 19.7 GB at 16-bit and 4.9 GB at 4-bit: the weights (8.2B parameters) plus a working margin. A long context needs more.

### What is the cheapest GPU to run Qwen3-8B 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.

### Can I use Qwen3-8B commercially?

Yes. Qwen3-8B is released under Apache License 2.0. The Apache License 2.0 is a permissive open-source license. It permits commercial use, modification and redistribution. It requires keeping the license and copyright notices and any NOTICE file, stating significant changes, and it includes an express patent grant from contributors.

### What is Qwen3-8B's context length?

40,960 tokens, from the maximum position embeddings in its published configuration.

## Similar Models

Model · Text generation

### [Qwen3-8B-AWQ](https://savrn.com/models/qwen3-8b-awq)

[Qwen](https://savrn.com/model-publishers/qwen)

Qwen3 is the latest generation of large language models in Qwen series, offering a comprehensive suite of dense and mixture-of-experts (MoE) models. Built upon extensive training, Qwen3 delivers groundbreaking advancements in reasoning, instruction-following, agent capabilities, and multilingual support, with the following key features: - Uniquely support of seamless switching between thinking mode (for complex logical reasoning, math, and coding) and non-thinking mode (for efficient, general-purpose dialogue) within single model, ensuring optimal performance across various scenarios. - Significantly enhancement in its reasoning capabilities, surpassing previous QwQ (in thinking mode) and…

Open weights apache-2.0 8.2B parameters 40,960 tokens transformers

[View model](https://savrn.com/models/qwen3-8b-awq)

Model · Text generation

### [DeepSeek-R1-0528-Qwen3-8B](https://savrn.com/models/deepseek-r1-0528-qwen3-8b)

[DeepSeek](https://savrn.com/model-publishers/deepseek-ai)

The DeepSeek R1 model has undergone a minor version upgrade, with the current version being DeepSeek-R1-0528. In the latest update, DeepSeek R1 has significantly improved its depth of reasoning and inference capabilities by leveraging increased computational resources and introducing algorithmic optimization mechanisms during post-training. The model has demonstrated outstanding performance across various benchmark evaluations, including mathematics, programming, and general logic. Its overall performance is now approaching that of leading models, such as O3 and Gemini 2.5 Pro. Compared to the previous version, the upgraded model shows significant improvements in handling complex reasoning…

Open weights mit 8.2B parameters 131,072 tokens transformers

[View model](https://savrn.com/models/deepseek-r1-0528-qwen3-8b)

Model · Text generation

### [Symbiotic-8B](https://savrn.com/models/symbiotic-8b)

[Convergent Intelligence](https://savrn.com/model-publishers/reaperdoesntknow)

Purpose: Long-memory symbolic reasoning + high-fidelity language generation SymbioticLM-8B is a state-of-the-art hybrid transformer model with built-in symbolic cognition. It combines an 8B Qwen-based transformer with modular symbolic processors and a persistent memory buffer. The model supports both general conversation and deep symbolic tasks such as theorem generation, logical chaining, and structured reasoning with retained memory across turns. - General symbolic reasoning and logical conversation - Code + math proof modeling - Not instruction-tuned (e.g., chat-style inputs may require prompt engineering) - Larger memory buffer may increase CPU load slightly - Symbolic inference is…

Open weights afl-3.0 8.2B parameters 40,960 tokens transformers

[View model](https://savrn.com/models/symbiotic-8b)

Model · Text generation

### [Qwen3-8B-CC-SFT-v2](https://savrn.com/models/qwen3-8b-cc-sft-v2)

[Liangzhidanta](https://savrn.com/model-publishers/liangzhidanta)

English | 简体中文 Qwen3-8B-CC-SFT-v2 is a failure-targeted continued-SFT checkpoint designed to improve coding-agent state preservation and continuation across Claude Code native context compaction. It is initialized from Qwen3-8B-CC-SFT-v1 and trained on compact-aware, context-correct supervision distilled from GLM-5.3 under real Claude Code native compaction. Best observed canonical303 Pass@1: 52.48% (159/303) — 4090 run of this same checkpoint at T=0.5 (temperature selected on an independent dev set; see the table below). Best observed configurations of this same checkpoint: - T=0.5 was selected on an independent 60-task development set (task-id and repo disjoint from canonical303), then…

Open weights apache-2.0 8.2B parameters 40,960 tokens transformers

[View model](https://savrn.com/models/qwen3-8b-cc-sft-v2)

Model · Text generation

### [ProactiveInquirer-Qwen3-8B-Merged](https://savrn.com/models/proactiveinquirer-qwen3-8b-merged)

[Ido Levy](https://savrn.com/model-publishers/dolev31)

Ido Levy 1,2 · 1,2 The trained questioner from Asking for What Was Never Requested: Horizontal and Vertical Proactivity in Agents, with its LoRA adapter merged into Qwen3-8B. It is a standard full-weight model: it loads without PEFT and serves with vLLM, SGLang or TGI like any Qwen3-8B. - The adapter, with the results, the training details, the limitations and a complete two-turn - Quantized for llama.cpp, Ollama and LM Studio: This is training seed 1, the adapter at the root of the adapter repository. The merge ran in float32 and the weights are stored in bfloat16. On the adapter card's two-turn example, greedy decoding with this model returns the adapter's output character for character.…

Open weights apache-2.0 8.2B parameters 40,960 tokens transformers

[View model](https://savrn.com/models/proactiveinquirer-qwen3-8b-merged)

Model · Text generation

### [r0_96](https://savrn.com/models/r0-96)

[Yonghoon](https://savrn.com/model-publishers/fiveflow)

Open weights apache-2.0 8.2B parameters 32,768 tokens transformers

[View model](https://savrn.com/models/r0-96)

## Qwen

[All models and datasets](https://savrn.com/model-publishers/qwen)

## Versions

- [b968826d9c46](https://savrn.com/models/qwen3-8b/versions/b968826d9c46) · current 2026-10-22

## Explore More

- [All text generation models](https://savrn.com/models/tasks/text-generation)
- [All models under apache-2.0](https://savrn.com/models/licenses/apache-2-0)
- [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.
- [Qwen website](https://qwen.ai/)
- [Hugging Face record](https://huggingface.co/Qwen/Qwen3-8B)
- [How the hub is built](https://savrn.com/model-hub/methodology)