# SambaNova SambaNova provides access to MiniMax M2.7 via the Anthropic Messages API on SambaCloud and demonstrated the model running on SambaRack SN50 with RDU chips at RAISE Summit 2026. SambaNova also joined the DOE-led Genesis Mission Consortium, contributing its purpose-built RDU technology. Status: provisional Last verified: 2026-08-31T14:13:22.000Z ## Evidence - (single_source) A single SambaRack SN50 contains 16 RDU chips. - supports https://sambanova.ai/blog/sn50-runs-fastest-minimax-speeds-in-the-world > a single SambaRack SN50 with 16 RDU chips - (single_source) The Responses API (/v1/responses) support starts with gpt-oss-120b, MiniMax M2.5, and MiniMax M2.7. - supports https://sambanova.ai/blog/build-faster-coding-agents-with-sambanovas-responses-api > /v1/responses support starts with gpt-oss-120b, MiniMax M2.5, and MiniMax M2.7. - (single_source) The SambaNova platform can run both large numbers of models and large models on a single system. - supports https://sambanova.ai/blog/sambanova-vs-cerebras > The SambaNova platform can run both large numbers of models and large models on a single system. - (single_source) MiniMax M2.7 is available via the Messages API on SambaCloud. - supports https://sambanova.ai/blog/sambacloud-now-supports-the-anthropic-messages-api > MiniMax M2.7 is also available via the Messages API on SambaCloud today. - (single_source) SambaNova is launching support for the Responses API across SambaCloud, SambaStack, and SambaManaged. - supports https://sambanova.ai/blog/build-faster-coding-agents-with-sambanovas-responses-api > SambaNova is launching support for the Responses API across the SambaNova platform — SambaCloud, SambaStack, and SambaManaged - (single_source) The base URL for using the Anthropic Messages API with SambaNova is https://api.sambanova.ai. - supports https://sambanova.ai/blog/sambacloud-now-supports-the-anthropic-messages-api > export ANTHROPIC_BASE_URL="https://api.sambanova.ai" - (single_source) The Messages API can be used within Claude Code by setting three environment variables: base URL, API key, and model name. - supports https://sambanova.ai/blog/sambacloud-now-supports-the-anthropic-messages-api > The Messages API can be used within Claude Code by setting three environment variables: base URL, API key, and model name. - (single_source) Server-side tools, PDF document blocks, and image URLs are not supported by the Messages API on SambaCloud. - supports https://sambanova.ai/blog/sambacloud-now-supports-the-anthropic-messages-api > Limitations to note: Server-side tools, PDF document blocks, and image URLs are not supported. - (single_source) SambaNova has joined the Genesis Mission Consortium, a public-private partnership led by the U.S. Department of Energy (DOE). - supports https://sambanova.ai/blog/sambanova-joins-the-genesis-mission-consortium > SambaNova has announced that it has joined the Genesis Mission Consortium, a landmark public-private partnership created to advance the Genesis Mission, which is led by the U.S. Department of Energy (DOE). - (single_source) The SambaRack SN50 demo uses one NVIDIA H200 rack with four GPUs for prefill and one SambaRack SN50 with 16 RDU chips for decode. - supports https://sambanova.ai/blog/sn50-runs-fastest-minimax-speeds-in-the-world > one NVIDIA H200 rack using four GPUs for prefill and one SambaRack SN50 with 16 RDU chips for decode. - (single_source) Teams can run an all-SambaNova stack with DeepSeek-V3.1 as planner and MiniMax M2.7 as executor under a single API key and billing relationship. - supports https://sambanova.ai/blog/build-faster-coding-agents-with-sambanovas-responses-api > Teams can run an all-SambaNova stack (DeepSeek-V3.1 as planner, MiniMax M2.7 as executor) under a single API key and billing relationship. - (single_source) Cerebras relies on SRAM only and is forced to commit large volumes of hardware to run even a single small model. - supports https://sambanova.ai/blog/sambanova-vs-cerebras > Cerebras, which relies on SRAM only, is forced to commit large volumes of hardware to run even a single small model. - (single_source) At RAISE Summit 2026, SambaNova demonstrated SambaRack SN50 running the fastest MiniMax M2.7 in a heterogeneous, disaggregated inference setup. - supports https://sambanova.ai/blog/sn50-runs-fastest-minimax-speeds-in-the-world > At RAISE Summit 2026, SambaNova is showing the next preview of premium inference: SambaRack SN50 running the fastest MiniMax M2.7 in a heterogeneous, disaggregated inference setup with one NVIDIA H200 rack using four GPUs for prefill and one SambaRack SN50 with 16 RDU chips for decode. - (single_source) The SambaNova SN40L Reconfigurable Dataflow Unit (RDU) is an AI accelerator engineered for large-scale AI inference with enterprise-level scalability. - supports https://sambanova.ai/blog/sambanova-vs-cerebras > The SambaNova SN40L Reconfigurable Dataflow Unit (RDU) is a cutting-edge AI accelerator, engineered to meet the demands of large-scale AI inference with enterprise-level scalability. - (single_source) The SN50 demo builds on the COMPUTEX blueprint that used NVIDIA B200 GPUs for prefill and SambaNova SN40 RDUs for decode. - supports https://sambanova.ai/blog/sn50-runs-fastest-minimax-speeds-in-the-world > The demo builds on the COMPUTEX blueprint we showed live with NVIDIA B200 GPUs for prefill and SambaNova SN40 RDUs for decode - (single_source) Codex CLI, Cline, and OpenCode all support the Responses API shape and can connect to SambaNova directly. - supports https://sambanova.ai/blog/build-faster-coding-agents-with-sambanovas-responses-api > Codex CLI, Cline, and OpenCode all support the Responses API shape and can connect to SambaNova directly. - (single_source) The recommended pattern for the Responses API is planner/executor: a high-reasoning model for planning and MiniMax M2.7 for fast, high-volume execution. - supports https://sambanova.ai/blog/build-faster-coding-agents-with-sambanovas-responses-api > The recommended pattern is planner/executor: a high-reasoning model for planning, MiniMax M2.7 for fast, high-volume - (single_source) The Genesis Mission Consortium was created to advance the Genesis Mission, focused on scientific, security, and energy challenges. - supports https://sambanova.ai/blog/sambanova-joins-the-genesis-mission-consortium > a landmark public-private partnership created to advance the Genesis Mission... challenges that will shape the nation's scientific, security, and energy future. - (single_source) SambaNova contributes its proprietary Reconfigurable Dataflow Unit (RDU) chip, purpose-built for AI inference rather than GPU-based training, to the Genesis Mission Consortium. - supports https://sambanova.ai/blog/sambanova-joins-the-genesis-mission-consortium > SambaNova contributes its proprietary Reconfigurable Dataflow Unit (RDU) chip, purpose-built for AI inference rather than GPU-based training. ## Timeline - 2026-08-31T14:03:51.000Z: MiniMax M2.7 is available through the Messages API on SambaCloud. The Messages API can be used in Claude Code by setting base URL, API key, and model name environment variables. Server-side tools, PDF document blocks, and image URLs are unsupported by the Messages API on SambaCloud. The base URL for the Anthropic Messages API with SambaNova is https://api.sambanova.ai. SambaNova joined the DOE-led Genesis Mission Consortium public-private partnership. SambaNova contributes its RDU chip, built for AI inference rather than GPU-based training, to the Genesis Mission Consortium. The Genesis Mission Consortium was created to advance scientific, security, and energy challenges. At RAISE Summit 2026, SambaNova demonstrated SambaRack SN50 running the fastest MiniMax M2.7 in a heterogeneous inference setup. The SambaRack SN50 demo uses an NVIDIA H200 rack with four GPUs for prefill and one SN50 with 16 RDU chips for decode. The SN50 demo builds on a COMPUTEX blueprint using NVIDIA B200 GPUs for prefill and SambaNova SN40 RDUs for decode. A single SambaRack SN50 contains 16 RDU chips. The SambaNova platform can run large numbers of models and large models on a single system.