首次启动会自动下载 Qwen3.8-27B 和 EAGLE-4 草稿模型，约 62 GB；用 pip 安装需要 Python 3.12，并额外准备 CUDA toolkit 和 C++ 编译器；启动后的前几个请求要编译 kernel 并实测开销，测速前先发几个请求预热；该软件是 source-available 许可，并非 OSI 开源，商业使用有营收门槛等限制。
docker run --gpus all --ipc host -p 127.0.0.1:30000:30000 -e HF_HOME=/hf -v ~/.cache/huggingface:/hf ghcr.io/eagle-hq/eagle-4-sglang:0.1.3 sglang serve --model-path Qwen/Qwen3.8-27B --host 0.0.0.0 --disable-overlap-schedule --speculative-algorithm EAGLE_HARNESS --speculative-draft-model-path EAGLE-Corp/Qwen3.8-27B-EAGLE4 --speculative-num-steps 12 --speculative-eagle-topk 4 --speculative-num-draft-tokens 16 --speculative-algorithm-config '{}'
pip install "eagle-harness[sglang]" "sglang==0.5.23.dev252+eagle.9" --extra-index-url https://eagle-hq.github.io/EAGLE-4/whl/cu130/ --extra-index-url https://download.pytorch.org/whl/cu130 --extra-index-url https://pypi.nvidia.com 硬件要求按 README 说明
快速开始测试于单张 NVIDIA B300（Linux + CUDA 13 驱动），实测加速用 4× B300；AMD MI355X 上 SGLang 也支持。README 未说明 CPU 运行方式。
需要 CUDA支持 ROCm / AMD有 Docker 镜像
EAGLE-4 是官方推理代码，用投机解码加速 LLM 生成，并自动适配硬件、批大小和工作负载
EAGLE-4 的官方推理代码，用投机解码加速大模型生成，可接入 SGLang、vLLM 等主流推理框架，显著提升 token 吞吐。
核心能力
- 用 eagle-harness 按实测开销决定投机深度
- 草稿树随批大小、硬件和工作负载自适应
- 输出与目标模型单独生成的结果完全一致
- 接入 SGLang、vLLM 和 TokenSpeed 三种引擎
- Qwen3.8-27B 上单请求最高 3.4× 加速
容易踩的坑
内容由模型根据 README 生成，可能不完整或不准确，以官方文档为准。看原文 README
- v0.1.5最新正式版2026-11-01 · 2026-11-01 02:14 UTC
EAGLE-4 combines a draft model with
eagle-harnessinside your inference engine to adapt speculative decoding to your hardware, batch and workload. Version 0.1.5 improves sampled decoding and startup on SGLang. Sampled draft trees now propose tokens from the request's own temperature, top-k, top-p and min-p settings, so more drafted tokens are accepted when you sample. The server also builds more of its kernels before it serves the first request.- Sampled trees follow the request's sampling settings. The draft tree draws its proposals with the request's temperature, top-k, top-p and min-p. It applies those filters in the engine's order.
- Draining batches keep their prices. When a batch shrinks to a size with no measured cost yet, live costs price it from the nearest measured sizes.
- More work before the first request. The server compiles the round tail's kernels and the native-decode position kernel at startup.
- Remaining requests keep their drafts. When some requests in a batch finish, the requests that remain keep their prelaunched draft trees. The server captures CUDA graphs for the round tail only after a batch size recurs.
- Draft attention splits by batch size. The draft's attention picks its key splits from the batch's size.
Every change is in the SGLang path of
eagle-harness. vLLM and TokenSpeed serving are unchanged, and the binary wheel carries the same native libraries as 0.1.4. The engine pins don't change: SGLang CUDA0.5.23.dev252+eagle.9, vLLM0.32.0rc1.dev283+eagle.5and TokenSpeed0.1.1+eagle.8.Install
You don't need to download anything from the release page. In a Python 3.12 environment, install the SGLang packages together:
pip install "eagle-harness[sglang]==0.1.5" "sglang==0.5.23.dev252+eagle.9" \ --extra-index-url https://eagle-hq.github.io/EAGLE-4/whl/cu130/ \ --extra-index-url https://download.pytorch.org/whl/cu130 \ --extra-index-url https://pypi.nvidia.comTo upgrade a 0.1.4 environment, run
pip install "eagle-harness==0.1.5"in it, then runeagle-harness-check. Nothing in the configuration changed. The install guide lists the vLLM and TokenSpeed pins, and the release images stay at:0.1.3with eagle-harness 0.1.3 installed.What was checked
We served Qwen3.8-27B with the EAGLE-4 draft on SGLang with the high-load recipe on four B300 GPUs. We compared 0.1.5 with 0.1.4 and with the same engine without speculation, using interleaved server starts. There were three starts each for 0.1.4 and plain decoding, and two completed starts for 0.1.5 (see Limits below).
-
Greedy speed. 0.1.5 measured 1.005× the throughput of 0.1.4 at 1 request, 0.997× at 8 and 0.983× at 64. That is 718 against 714 output tokens per second at 1 request, 2,918 against 2,927 at 8, and 8,003 against 8,142 at 64. Each 0.1.5 mean is inside the range of the 0.1.4 starts. At 64 requests the recipe decodes without speculation in both versions.
-
Sampling acceptance. At temperature 1, the mean accepted length per round rose:
Sampling Requests 0.1.4 0.1.5 top-k 8, top-p 0.7 1 4.30 4.55 top-k 8, top-p 0.7 8 3.88 4.16 min-p 0.05 1 4.12 4.17 min-p 0.05 8 3.81 3.99 no filter 1 4.01 4.04 no filter 8 3.84 3.86 Sampled throughput measured 0.98× to 1.01× that of 0.1.4, and each 0.1.5 mean is inside the range of the 0.1.4 starts, apart from one cold-cache start described under Limits.
-
Greedy correctness. On 20 prompts of 512 tokens at 1 request, the 0.1.5 starts produced output identical to the first 0.1.4 start on 20 of 20 and 18 of 20 prompts. The two 0.1.4 repeat starts matched it on 20 and 19 of 20, and one of the 0.1.5 differences is at the same token of the same prompt as the 0.1.4 one. Eve
- v0.1.42026-10-28 · 2026-10-28 20:17 UTC
查看更新说明
EAGLE-4 combines a draft model with
eagle-harnessinside your inference engine to adapt speculative decoding to your hardware, batch and workload. Version 0.1.4 makes vLLM serving on two or more GPUs more robust. During server startup and CUDA graph capture, vLLM issues its symmetric-memory collectives from several CUDA streams, and CUDA does not order work across streams. This release orders those collectives, so each all-reduce and each candidate all-gather runs alone on its buffer on every GPU, and startup and graph capture complete with correct results.- vLLM's symmetric-memory all-reduce (on by default,
VLLM_ALLREDUCE_USE_SYMM_MEM=1) now waits for the previous all-reduce whenever the calling stream changes. - The
eagle-harnesscandidate all-gather on vLLM does the same.
Calls from one stream were already ordered, so steady-state decoding does no extra work. One-GPU vLLM, SGLang and TokenSpeed do not use these paths.
This release ships vLLM
0.32.0rc1.dev283+eagle.5, built from eagle-hq/vllmeagle/v0.1.4. It is+eagle.4plus the all-reduce fix, with the same native libraries. SGLang CUDA0.5.23.dev252+eagle.9and TokenSpeed0.1.1+eagle.8are unchanged.Install
You don't need to download anything from the release page. In a Python 3.12 environment, install the vLLM packages together:
pip install "eagle-harness[vllm]==0.1.4" "vllm==0.32.0rc1.dev283+eagle.5" \ --extra-index-url https://eagle-hq.github.io/EAGLE-4/whl/cu130/ \ --extra-index-url https://download.pytorch.org/whl/cu130To upgrade a 0.1.3 vLLM environment, run this command, then run
eagle-harness-check. SGLang and TokenSpeed users can installeagle-harness==0.1.4with their 0.1.3 engine pins: the install guide lists them. Nothing in the configuration changed.Engine 0.1.4 pins SGLang CUDA eagle-harness[sglang]==0.1.4,sglang==0.5.23.dev252+eagle.9(unchanged)vLLM eagle-harness[vllm]==0.1.4,vllm==0.32.0rc1.dev283+eagle.5TokenSpeed eagle-harness[tokenspeed]==0.1.4,tokenspeed==0.1.1+eagle.8and its 0.1.3 scheduler, kernel and engine wheels (unchanged)What was checked
- Stress tests on B300. Collectives alternated between two calling streams while one GPU lagged 20 ms on every seventh call, and every result was checked:
- All-reduce: 40 of 40 check windows returned the expected sums at both 8 and 2 GPUs.
- Candidate all-gather: 400 of 400 results matched the expected rows at both 8 and 2 GPUs.
- Serving check. We served Qwen3.8-27B with the EAGLE-4 draft on vLLM and compared the fixed build with 0.1.3 on four B300 GPUs and on one, with greedy decoding. That build also enabled three default-off options that are not part of this release, and they changed no output:
- Outputs: token-identical between the two builds on all 20 prompts, at both GPU counts. Both builds pass the release correctness check against vLLM without speculation.
- Logprobs: identical on all 768 checked positions.
- Speed: four interleaved server starts per build, at 1, 32, 64 and 128 concurrent requests. Every cell is within the start-to-start spread; the fixed build measures 0.99× to 1.04× the throughput of 0.1.3.
- Measured throughput: 713.6 against 712.6 output tokens per second at 1 request on four GPUs, and 7,890 against 7,858 at 64.
- Release packages. A fresh environment installed from the exact 0.1.4 wheels served the same 20 prompts on one B300 with token-identical output and identical logprobs, and the candidate all-gather GPU tests passed.
Limits
This release changes nothing else, so the 0.1.3 limits and known limits still apply. Two items that the 0.1.3 notes planned for 0.1.4 move to a l
- vLLM's symmetric-memory all-reduce (on by default,
- v0.1.32026-10-27 · 2026-10-27 03:27 UTC
查看更新说明
EAGLE-4 combines a draft model with
eagle-harnessinside your inference engine to adapt speculative decoding to your hardware, batch and workload. Version 0.1.3 brings native decoding at high load to the SGLang README command, including the one-GPU setup, and fills draft KV without prediction work for prompts admitted above the verify limit. It also builds engine wheels and release images from approved inputs and explains how to warm up a server before measuring it.This release ships SGLang CUDA
0.5.23.dev252+eagle.9, keeps vLLM0.32.0rc1.dev283+eagle.4unchanged, and adds TokenSpeed0.1.1+eagle.8, built from eagle-hq/tokenspeedeagle/v0.1.3. The TokenSpeed engine exposes generic engine hooks, and EAGLE-4 policy and tree verification live in the harness plugin. TokenSpeed serves EAGLE-4 in serve mode from cost tables you calibrate once on your setup. On four B300 GPUs this build passed calibration, greedy serving from the calibrated profile (twenty prompts: 3 identical to TokenSpeed without speculation, 17 first divergences within the fp32 tolerance, the same in two starts) and sampled serving at temperature 1. TokenSpeed's 0.1.2 speed measurements remain historical results, not measurements of this build.Install
You don't need to download anything from the release page. In a Python 3.12 environment, install the SGLang packages together:
pip install "eagle-harness[sglang]==0.1.3" "sglang==0.5.23.dev252+eagle.9" \ --extra-index-url https://eagle-hq.github.io/EAGLE-4/whl/cu130/ \ --extra-index-url https://download.pytorch.org/whl/cu130 \ --extra-index-url https://pypi.nvidia.comFor TokenSpeed, install its runtime with the matching scheduler, kernel and engine wheels:
pip install "eagle-harness[tokenspeed]==0.1.3" "tokenspeed==0.1.1+eagle.8" \ "tokenspeed-scheduler==0.1.42+eagle.8" "tokenspeed-kernel==0.1.4+eagle.8" \ "tokenspeed-engine==0.1.0+eagle.8" \ --extra-index-url https://eagle-hq.github.io/EAGLE-4/whl/cu130/ \ --extra-index-url https://download.pytorch.org/whl/cu130Follow the SGLang guide, the vLLM guide or the TokenSpeed guide to serve an EAGLE-4 draft. The TokenSpeed guide calibrates cost tables once, then serves in serve mode. The TokenSpeed image is
:0.1.3. The SGLang and vLLM:0.1.2images remain the documented images until the:0.1.3image checks finish.Measured speed
These measurements use Qwen3.8-27B with the EAGLE-4 draft on NVIDIA B300 GPUs and greedy decoding. Each cell is EAGLE-4 output throughput divided by the same engine's throughput without speculation, averaged over four starts per configuration. Client concurrency ranges from 1 to 512 for the README and high-load recipes, and from 1 to 64 for the measured recipes. Controls use the same recipe, tensor-parallel size and running-request cap. These throughput results do not imply a passing correctness check, as described below.
Engine and command GPUs 1 request 8 64 128 256 512 SGLang README 1 2.93× 1.99× 1.00× 1.00× 1.00× 0.99× SGLang README 4 2.34× 1.64× 0.99× 1.00× 1.02× 1.03× SGLang measured recipe 4 3.37× 2.06× 1.27× Not measured Not measured Not measured SGLang high-load recipe 4 2.90× 1.81× 1.00× 0.98× 1.01× 0.99× vLLM measured recipe 4 2.91× 1.74× 1.00× Not measured Not measured Not measured The corresponding EAGLE-4 throughput is in output tokens per second:
Engine and command GPUs 1 request 8 64 128 256 512 SGLang README 1 309.5 1,356.6 3,336.8 - v0.1.22026-10-23 · 2026-10-23 02:00 UTC
查看更新说明
EAGLE-4 0.1.2 updates
eagle-harnesson the 0.1.1 engines. On SGLang, EAGLE-4 decodes a batch without speculation wherever speculation would be slower, so the high-load recipe keeps pace with plain decoding up to 512 concurrent requests on four B300 GPUs and keeps its speedup at low load.eagle-harness calibrate --engine vllmcompletes on hybrid targets, and the TokenSpeed draft runs its candidate head on the engine's low-latency GEMM.Install
You don't need to download anything from this page.
pipfetches everything:eagle-harnesscomes from PyPI and the engines come from our wheel index. Install one engine per environment, with the plugin:python3.12 -m venv .venv && source .venv/bin/activate pip install "eagle-harness[sglang]" "sglang==0.5.23.dev252+eagle.5" \ --extra-index-url https://eagle-hq.github.io/EAGLE-4/whl/cu130/ \ --extra-index-url https://download.pytorch.org/whl/cu130 \ --extra-index-url https://pypi.nvidia.comTo upgrade a 0.1.1 environment, run the same command with
"eagle-harness[sglang]==0.1.2". The installation guide has the vLLM, TokenSpeed and ROCm steps. For high load, serve with the high-load recipe.Measured throughput
4× B300, TP4, temperature 0. The speedup divides EAGLE-4 throughput by the same engine with speculation off, measured in the same allocation.
Engine and command Concurrency No-spec tok/s EAGLE-4 tok/s Speedup Starts SGLang, measured recipe 1 219.0 733.8 3.35× 2 SGLang, measured recipe 8 1545.2 3257.9 2.11× 2 SGLang, measured recipe 64 4718.6 6488.2 1.38× 2 SGLang, high-load recipe 64 7313.7 7393.3 1.01× 3 SGLang, high-load recipe 128 11086.4 11114.3 1.00× 3 SGLang, high-load recipe 256 14264.9 14332.3 1.00× 3 SGLang, high-load recipe 512 16340.7 16197.5 0.99× 3 vLLM, measured recipe 1 239.4 678.5 2.83× 4 vLLM, measured recipe 8 1668.9 2887.4 1.73× 4 vLLM, measured recipe 64 6725.4 8013.4 1.19× 4 TokenSpeed, measured recipe 1 283.5 886.5 3.13× 2 TokenSpeed, measured recipe 8 1628.2 3256.4 2.00× 2 TokenSpeed, measured recipe 64 4969.8 6974.1 1.40× 2 The high-load rows are restarts with warm caches; a first start on a machine measured 0.95× at 64 and 0.97× at 128. Every start also passes the twenty-prompt greedy correctness check. The release notes have the conditions.
Limits in this release
- The SGLang quick-start command is slower than plain decoding above about 32 to 64 concurrent requests; for high load, use the high-load recipe on four GPUs. A fix is planned.
- High load on one GPU is not supported yet.
- vLLM and TokenSpeed do not switch to plain decoding at high load yet; above concurrency 64, serve them without speculation for now.
- Known limits lists the rest.
Files
No files on this page. The engine wheels are on the v0.1.0 release, which our wheel index serves, and
eagle-harnessis on PyPI. The:0.1.2images (ghcr.io/eagle-hq/eagle-4-sglang:0.1.2,ghcr.io/eagle-hq/eagle-4-vllm:0.1.2,ghcr.io/eagle-hq/eagle-4-tokenspeed:0.1.2) carryeagle-harness0.1.2. - v0.1.12026-10-21 · 2026-10-21 16:23 UTC
查看更新说明
EAGLE-4 0.1.1 updates
eagle-harnessand two engine builds. An SGLang server now needs only the draft path to serve EAGLE-4 and starts its first requests faster, and a new SGLang CUDA engine adds rounds without speculation where the live costs pick them and a temperature-1 sampled tree. vLLM measures the tree acceptance it serves, and the TokenSpeed integration moves to a new engine build. The vLLM engine is unchanged.Install
You don't need to download anything from this page.
pipfetches everything:eagle-harnesscomes from PyPI and the engines come from our wheel index. Install one engine per environment, with the plugin:python3.12 -m venv .venv && source .venv/bin/activate pip install "eagle-harness[sglang]" "sglang==0.5.23.dev252+eagle.5" \ --extra-index-url https://eagle-hq.github.io/EAGLE-4/whl/cu130/ \ --extra-index-url https://download.pytorch.org/whl/cu130 \ --extra-index-url https://pypi.nvidia.comTo upgrade a 0.1.0 environment, run the same command with
"eagle-harness[sglang]==0.1.1". The installation guide has the vLLM, TokenSpeed and ROCm steps.Breaking: a config that still sets
"mode": "eval",ragged,cost_model.ragged_*ordraft_model.moe_gate_accstops at load. Delete the option; the migration guide lists each one.Measured throughput
4× B300, TP4, temperature 0. The speedup divides EAGLE-4 throughput by the same engine with speculation off, measured in the same allocation.
Engine Concurrency No-spec tok/s EAGLE-4 tok/s Speedup Starts SGLang 1 218.5 750.6 3.44× 3 SGLang 8 1541.7 3203.1 2.08× 3 SGLang 64 4655.3 6454.1 1.39× 3 SGLang, 32K context 1 239.7 715.3 2.98× 4 vLLM 1 240.5 681.0 2.83× 6 vLLM 8 1686.7 2894.8 1.72× 6 vLLM 64 6950.5 7935.3 1.14× 6 Every start also passes the twenty-prompt greedy correctness check. The release notes have temperature 1 and the conditions.
Limits in this release
- SGLang above concurrency 64 is slower with speculation than without at a running cap of 512; serve those loads without speculation for now.
- TokenSpeed runs on NVIDIA CUDA only.
- Known limits lists the rest.
Files
The images are
ghcr.io/eagle-hq/eagle-4-sglang:0.1.1,ghcr.io/eagle-hq/eagle-4-vllm:0.1.1andghcr.io/eagle-hq/eagle-4-tokenspeed:0.1.1. No other files on this page. The new engine wheels are on the v0.1.0 release, which our wheel index serves, andeagle-harnessis on PyPI.
用户点评
登录用户的真实使用记录，可按版本和硬件筛选共 0 条
还没有人点评。第一个说出真实体验的人最有价值。
EAGLE-4
Leave speculation on. EAGLE-4 adapts to your hardware, batch and workload.
For high load, serve SGLang on four B300 GPUs with the high-load recipe, which keeps pace with plain decoding up to 512 concurrent requests.
EAGLE-4 (Extrapolation Algorithm for Greater Language-model Efficiency) speeds up LLM inference with
speculative decoding. EAGLE-4 draft models are built to run with eagle-harness, which decides how far to
speculate from measured costs on your setup, so every draft tree fits the batch size, the hardware and the
workload. The output stays exactly what the target model would generate on its own: up to 3.4× faster
generation on Qwen3.8-27B, with the same answers. Measured speedups
This repository holds eagle-harness, the package that brings EAGLE-4 to SGLang, vLLM and TokenSpeed.
Install it, point your engine at an EAGLE-4 draft model, and it serves with EAGLE-4.
Quick start
The quick start serves on one GPU and is tested on one NVIDIA B300 with Linux and a CUDA 13 driver.
Qwen3.8-27B and its
EAGLE-4 draft download on the first start (about 62 GB).
The measured speedups use four B300 GPUs. Add --tp-size 4 to either serve command to serve
on four, and see the engine guides for the exact measured commands.
Start an SGLang server with Docker:
docker run --gpus all --ipc host -p 127.0.0.1:30000:30000 \
-e HF_HOME=/hf -v ~/.cache/huggingface:/hf \
ghcr.io/eagle-hq/eagle-4-sglang:0.1.3 \
sglang serve --model-path Qwen/Qwen3.8-27B --host 0.0.0.0 \
--disable-overlap-schedule --speculative-algorithm EAGLE_HARNESS \
--speculative-draft-model-path EAGLE-Corp/Qwen3.8-27B-EAGLE4 \
--speculative-num-steps 12 --speculative-eagle-topk 4 --speculative-num-draft-tokens 16 \
--speculative-algorithm-config '{}'
Or install with pip, in Python 3.12. Pip installs also need the CUDA toolkit and a C++ compiler:
python3.12 -m venv .venv && source .venv/bin/activate
pip install "eagle-harness[sglang]" "sglang==0.5.23.dev252+eagle.9" \
--extra-index-url https://eagle-hq.github.io/EAGLE-4/whl/cu130/ \
--extra-index-url https://download.pytorch.org/whl/cu130 \
--extra-index-url https://pypi.nvidia.com
sglang serve --model-path Qwen/Qwen3.8-27B \
--speculative-algorithm EAGLE_HARNESS \
--speculative-draft-model-path EAGLE-Corp/Qwen3.8-27B-EAGLE4
When the server logs The server is fired up and ready to roll!, send a request from another terminal:
curl http://localhost:30000/v1/chat/completions -H 'Content-Type: application/json' \
-d '{"model": "Qwen/Qwen3.8-27B", "messages": [{"role": "user", "content": "Explain how a hash table works."}], "max_tokens": 256}'
The first requests after a start compile kernels and measure live costs, so they run slower. Send a few requests
before you measure speed, and compare against the same command without the --speculative-* flags
(benchmarking).
For vLLM, TokenSpeed and AMD MI355X, follow the engine guides.
Measured speedup
Each speedup divides EAGLE-4 throughput by the same engine's throughput without speculation, on Qwen3.8-27B at temperature 0 on the SPEED-Bench client, with the measured recipe of each engine guide. The support matrix has the tokens/s, the correctness checks and how each number was measured.
| Engine | GPUs | 1 request | 8 concurrent | 64 concurrent |
|---|---|---|---|---|
| SGLang | 4× B300 | 3.42× | 2.09× | 1.39× |
| vLLM | 4× B300 | 2.83× | 1.72× | 1.14× |
| TokenSpeed | 4× B300 | 3.09× | 2.02× | 1.41× |
On one AMD MI355X, SGLang with EAGLE-4 is 2.07× faster at 8 concurrent requests and 1.56× at 32.
More
- Install: pins, wheel indexes, Docker images and the ROCm closure.
- Engine guides with the measured recipes: SGLang, vLLM, TokenSpeed, AMD MI355X.
- Support matrix and known limits.
- Calibration, configuration and troubleshooting.
- How it works, kernels, compatibility, new models and the FAQ.
Citation
If you use EAGLE-4, cite the EAGLE-4 paper. EAGLE-4 builds on EAGLE, EAGLE-2 and EAGLE-3, which
CITATION.cff also lists.
@misc{eagle4,
author = {EAGLE-4 Team},
title = {{EAGLE-4}: One Flexible Drafter for All Batch Sizes and Hardware},
year = {2026},
note = {Preprint}
}
Acknowledgements
- Thanks to Aditya Nandakumar (@proaditya) and the contributors to the original EAGLE-4 PR 1 for the ROCm port of EAGLE-4 on AMD Instinct MI355X.
- Thanks to Andy Luo (@andyluo7) and the contributors to the original EAGLE-4 PR 2 for the AMD AITER and Triton kernels on GFX950.
- Thanks to Andy Luo (@andyluo7), Aditya Nandakumar (@proaditya) and the contributors to the merged AMD series in the original EAGLE-4 PR 4 for AITER/FA3 and Split-KV on GFX950.
- Thanks to Nicolas Perrin (@nperrin-fr) and the TokenSpeed/LightSeek contributors for the TokenSpeed implementation in the original EAGLE-4 PR 3, including the MLP L2 warm-ahead and TGV SwiGLU work used in the TokenSpeed fork.
- SPEED-Bench is by NVIDIA. Model weights are licensed by their owners.
License and trademark policy
Eagle 4 is source-available software, not OSI-approved open-source software. The terms of the Eagle 4 Source-Available License govern its use, modification, and distribution, and incorporate the Eagle Trademark Policy. See also the NOTICE file.
Commercial Use of a Competing Offering requires a separate written commercial license regardless of revenue. Other Commercial Use is permitted only within the conditions of the license, including the US $20 million Aggregate Revenue threshold measured over any consecutive 12-month period. If the threshold is exceeded, a separate commercial agreement is required before beginning or continuing any Commercial Use. The license specifies that the requirement continues after a subsequent revenue decrease.
Downloading, cloning, forking, copying, accessing, modifying, using, hosting, distributing, or otherwise exercising any licensed right constitutes acceptance of the Source-Available License and incorporated Eagle Trademark Policy. Mere viewing or inspection does not itself constitute trademark use.
For separate commercial terms, see the
Commercial License Agreement and contact
[email protected]. Publishing that agreement does
not grant commercial-license rights; it becomes binding only as stated in
its Section 1. Files derived from
sglang, vLLM, FlashInfer and other projects keep those projects' notices, listed in
THIRD_PARTY_NOTICES.md.
Contributing
By submitting, uploading, pushing, committing, proposing, or otherwise providing
a pull request, patch, issue attachment, model, weight, dataset, documentation,
or other contribution to this repository, you agree to the Eagle Contribution
License and Submission Terms
(PDF). If you do not agree, do not submit a contribution. If you act for an
employer, client, partner or other entity, you represent that you have authority
to submit the contribution and have obtained all required permissions.
Submitting a contribution does not change the license of existing Eagle code;
Eagle AI Corp controls outbound licensing for accepted contributions. See
CONTRIBUTING.md for the licensing structure and for how to
build, test and open a pull request.