Analysis updated 2026-07-07 · repo last pushed 2026-05-11
Compare the live speed of Nvidia versus AMD GPUs running popular AI models to decide which hardware to buy.
Track how a new version of inference software like vLLM improves token throughput day over day.
View a dashboard showing performance per dollar and energy efficiency for different chip and software combinations.
Plan multi-million dollar data center purchases using real-world benchmark data instead of theoretical specs.
| zihaomu/inferencex | 0xhassaan/nn-from-scratch | 0xzgbot/hermes-comfyui-skills | |
|---|---|---|---|
| Stars | — | 0 | 0 |
| Language | — | Python | — |
| Last pushed | 2026-05-11 | — | — |
| Maintenance | Maintained | — | — |
| Setup difficulty | hard | moderate | easy |
| Complexity | 4/5 | 4/5 | 1/5 |
| Audience | ops devops | developer | designer |
Figures from each repo's GitHub metadata at analysis time.
Requires enterprise-grade GPU hardware from major vendors and significant compute resources to run continuous, real-world AI model benchmarks.
InferenceX is a free, open-source benchmarking tool that continuously measures how fast different AI models run across various hardware chips. Instead of running a speed test once and publishing results that quickly become outdated, it constantly re-tests performance as software gets updated, giving you a live, near-real-time view of which combinations of hardware and software deliver the best results. When a company serves an AI model to users, the speed and efficiency of that model depend heavily on two things: the physical chips (like Nvidia or AMD GPUs) and the inference software (the programs that actually run the model). Because developers constantly release software updates that make these programs faster, sometimes just days apart, any benchmark from a few months ago is already obsolete. This project solves that problem by running automated, continuous tests to capture those incremental daily improvements, tracking real-world metrics like token throughput, performance per dollar, and energy efficiency. This project is built for data center operators, AI infrastructure teams, and hardware companies who need to make multi-million dollar decisions about which equipment and software to buy or use. For example, a cloud provider deciding whether to invest in a new fleet of Nvidia Blackwell chips or AMD MI355X GPUs could use the live dashboard to see exactly how those chips perform right now with popular software like vLLM or SGLang. It takes the guesswork out of infrastructure planning by showing how theoretical hardware specs actually translate to real-world AI speeds. The project is notable for its high level of industry backing. It has received physical hardware, compute resources, and technical support directly from the CEOs and engineering teams of major companies like AMD, Nvidia, and OpenAI, alongside various cloud providers. This corporate support ensures the tests are run on top-tier, enterprise-grade machines rather than standard cloud instances, making the results highly credible for large-scale infrastructure decision-making.
InferenceX is a free, open-source tool that continuously benchmarks AI model speed across different hardware chips, giving a live view of which combinations of hardware and software perform best.
Maintained — commit in last 6 months (last push 2026-05-11).
The explanation does not specify the license, so what it allows is unknown.
Setup difficulty is rated hard, with roughly 1day+ to a first successful run.
Mainly ops devops.
This repo across BitVibe Labs
Verify against the repo before relying on details.