RadixArk

Name: RadixArk
Brand: RadixArk

AI Native

BUILD AI FOR ALL

Est 2025|Valuation $400M|5 Ref Customers

LinkedIn 3K|Website|FTEs 42|HQ San Francisco, California, United States

01Competitive Landscape

Narrative analysis

RadixArk vs vLLM: The most direct comparison is at the serving layer (SGLang vs vLLM). Community discussions indicate SGLang can outperform in some time-to-first-token and throughput scenarios, particularly where prefix reuse helps, but vLLM may have advantages in ecosystem maturity and earlier support for new models.
RadixArk vs Fireworks/Baseten/Together: Those are managed inference platforms that tend to offer faster time-to-value and predictable operations. RadixArk’s potential advantage is combining an optimized serving engine with training/RL infrastructure and an open-source foundation.

- RadixArk vs vLLM: The most direct comparison is at the serving layer (SGLang vs vLLM). Community discussions indicate SGLang can outperform in some time-to-first-token and throughput scenarios, particularly where prefix reuse helps, but vLLM may have advantages in ecosystem maturity and earlier support for new models. - RadixArk vs Fireworks/Baseten/Together: Those are managed inference platforms that tend to offer faster time-to-value and predictable operations. RadixArk’s potential advantage is combining an optimized serving engine with training/RL infrastructure and an open-source foundation.

Adoption trends (caveat)

Press frames RadixArk as part of a trend of open-source inference projects becoming venture-backed companies, suggesting market interest.
No authoritative market-share numbers were found in accessible sources.

	RadixArk (managed + open source)	vLLM (open source)	Fireworks AI (hosted)
Starting Price	Not clearly public (sales-led implied)	Open-source software (infrastructure and operational costs apply)	Published usage-based pricing available from the vendor (varies by model); batch inference discounts noted
Best For	Teams wanting SGLang/Miles performance with optional managed services	Teams that want a widely adopted inference server and will self-operate	Teams that want hosted inference quickly with transparent usage billing
Key Strength	Open-source infrastructure approach and emphasis on cache/prefix reuse (via SGLang)	Broad ecosystem adoption, often fast model support, self-host control	Convenience, hosted operations, public pricing, managed scaling
Key Weakness	New commercial offering; limited public enterprise security/compliance artifacts; limited reviews	Requires infra/SRE investment; performance tuning burden; not a managed service	Vendor lock-in risk; may be pricier at scale than self-hosting; less customization than self-managed

RadixArk

01Competitive Landscape

Narrative analysis

Adoption trends (caveat)

02Market Map

01Competitive Landscape

Narrative analysis

Adoption trends (caveat)

02Market Map