SambaNova Systems

license

proprietary

source

closed(managed cloud API, the EULA states Customer has no right to obtain a copy of the underlying computer code for any Service)

software-stack

closed(SambaFlow compiler/runtime not open)

hardware

SN40L-RDU-only

delivery

managed-cloud-API

Proprietary managed cloud inference on SambaNova's own RDU silicon, with no open-source software disclosed. The Cloud EULA grants only a limited license to use the software embodied in the service and states that the customer has no right to obtain a copy of the underlying computer code, so there is no self-hostable implementation at all, which settles the score on source alone.

• https://sambanova.ai/blog/sn40l-chip-best-inference-solution recorded 2026-08-09

  The RDU is also the only AI accelerator with a tightly coupled three-tier memory system comprising SRAM, HBM, and DDR DRAM.

• https://sambanova.ai/cloud-end-user-license-agreement recorded 2026-09-23

  SambaNova hereby grants to Customer a non-exclusive, non-transferable, limited license during the Term to use the software embodied in the Service. The same agreement states: "Customer acknowledges that the Service is offered as an online, hosted solution, and that Customer has no right to obtain a copy of the underlying computer code for any Service", and forbids Customer to "reverse engineer, disassemble or decompile any Service or otherwise seek to obtain the source code of any software included in the Service".

OpenRouter's provider page for SambaNova charts daily tokens processed on the models routed to SambaNova there. SambaNova publishes no user or customer count, so this third-party traffic is the only usage measure found, and it covers only what is routed through OpenRouter.

• https://openrouter.ai/provider/sambanova recorded 2026-09-29

  OpenRouter provider page for SambaNova has a "Tokens processed on OpenRouter" chart and states "Access 7 models served by SambaNova through the OpenRouter unified API"; the embedded daily series for 2026-09-28 sums to about 1.07B tokens, led by openai/gpt-oss-120b at about 584M, and 2026-09-27 to about 0.75B.

• https://sambanova.ai/blog/sn40l-chip-best-inference-solution recorded 2026-09-29

  SambaNova blog offers free trial inference on SambaNova Cloud: "Try our blazing fast inference performance for yourself for free at" cloud.sambanova.ai, alongside its SN40L benchmark tables.

SambaNova's own SN40L benchmarks and Artificial Analysis's independent measurements both show strong per-token throughput across the models it serves. Its served-model catalog is narrower than the largest multi-tenant inference clouds, which is the constraint keeping it a step behind them.

• https://artificialanalysis.ai/providers/sambanova recorded 2026-09-29

  Artificial Analysis lists SambaNova with "9 models, each with different intelligence, performance, and pricing characteristics" and states "the fastest models are gpt-oss-120b (low) (727 t/s), gpt-oss-120b (high) (709 t/s), and MiniMax-M2.7 (397 t/s)"; the model table lists Llama 3.3 70B at 278 output tokens/s.

• https://sambanova.ai/blog/sn40l-chip-best-inference-solution recorded 2026-09-29

  SambaNova blog reports Llama 3.1 405B at "129 output tokens/second/user"; Table 1 lists the SambaNova 16-bit column at 457 for Llama 3.1 70B and 1042 for 8B.