Vast-10M (Voltropy)

Summary

Voltropy PBC introduces Vast-10M, billed as the first frontier LLM with a ten-million-token context window, in three sizes: Vast-10M-Flash (based on DeepSeek V4 Flash), Vast-10M-Medium (based on GLM 5.2), and Vast-10M-Pro (based on DeepSeek V4 Pro). The context extension technique is called VSA; per Voltropy, unlike prior context-extension methods that degrade intelligence, VSA improves models even at shorter context lengths. On the BEAM benchmark, Vast-10M-Flash beats Anthropic's Fable 5.1 at the one-million-token tier and reaches parity with OpenAI's GPT-6 Astra; Voltropy claims greater recall at ten million tokens than the base DeepSeek model at one million. The full technical report is available on voltropy.com; early access opened 2026-09-29.

Keywords

long context · 10M context · VSA · BEAM · models · agents

Topics

long context, language models, benchmarks

Research notes

• Discovery: Voltropy (@Voltropy, verified) on 2026-09-29: https://x.com/Voltropy/status/2104948250021085203
• Self-reply: early access open at voltropy.com; technical paper live.
• Press release: https://www.prnewswire.com/news-releases/voltropy-unveils-ai-that-can-process-10x-more-data-than-top-labs-302892256.html
• Voltropy PBC, Orange County CA; known for LCM (Lossless Context Management) integrated with OpenClaw and Hermes Agent after outperforming Anthropic's Claude Code on OOLONG.
• Connects to the collection's long-context and agent entries.