{"slug": "maple-said-it-made-deepseek-v4-flash-available-to-paid-customers", "title": "Maple said it made DeepSeek V4 Flash available to paid customers", "summary": "Maple, an Austin-based encrypted AI access company, announced it made DeepSeek V4 Flash available to paid customers, though its public materials do not specify qualifying plans, usage limits, or whether access includes Maple Research, Maple Proxy, or both. Co-founder and CEO Mark Suman, whose background includes AI and privacy work at Apple and roles at Instructure, positions Maple as a privacy layer around models developed elsewhere, with prompts encrypted on-device and decrypted inside hardware-isolated enclaves. DeepSeek's V4 Flash, introduced April 24, 2026, is a 284-billion-parameter mixture-of-experts model with 13 billion active parameters per token and a one-million-token context window.", "body_md": "[Maple](https://www.trymaple.ai/?ref=runtimewire) said in an undated announcement that it had made DeepSeek V4 Flash available to paid customers, but its public materials do not identify the qualifying plans, usage limits or whether access includes Maple Research, Maple Proxy or both.\n\n[Maple announcement on X](https://x.com/TryMapleAI/status/2088385094847254629?ref=runtimewire)\n\nCo-founder and CEO Mark Suman is building Maple, an Austin company centered on encrypted access to models developed elsewhere, offering users a way to run prompts without exposing plaintext to the model provider.\n\nIn its [Proxy announcement](https://blog.trymaple.ai/introducing-maple-proxy-the-maple-ai-api-that-brings-encrypted-llms-to-your-openai-apps/?ref=runtimewire), Maple said it was launching the service. The supplied public materials do not establish whether DeepSeek V4 Flash is available through Maple Proxy or provide model-specific usage limits and pricing.\n\nMaple's [current pricing page](https://trymaple.ai/pricing?ref=runtimewire) does not specify which subscriptions include V4 Flash. Maple lists Pro at $20 a month, Max at $100 a month and Team at $30 per user per month in its [getting-started guide](https://blog.trymaple.ai/getting-started-with-maple-ai-a-step-by-step-guide/?ref=runtimewire). Those published prices do not establish V4 Flash access or usage terms.\n\nSuman has worked in technology, including AI and privacy work at Apple and roles at Instructure, according to [Maple's founder page](https://www.trymaple.ai/about?ref=runtimewire).\n\nSuman has tied Maple's origin to a password leak he encountered roughly two decades ago. The incident became more consequential in the age of AI, he [wrote in January](https://blog.trymaple.ai/one-year-of-private-ai-how-maple-ai-made-encryption-easy/?ref=runtimewire), because users now put legal plans, business documents and personal thoughts into hosted systems. His founding thesis is direct: \"privacy shouldn't require sacrifice.\"\n\n### Maple sells a privacy layer around the model\n\nIn an [April 24, 2026 release post](https://api-docs.deepseek.com/news/news260424/?ref=runtimewire), DeepSeek introduced V4 Flash alongside the larger V4 Pro. DeepSeek describes V4 Flash as a 284-billion-parameter mixture-of-experts model that activates 13 billion parameters for each token. It supports a one-million-token context window, reasoning and non-reasoning modes, JSON output and tool calling.\n\nThose specifications could suit long documents, research sessions and agent workflows, where context capacity and inference cost can matter as much as benchmark performance.\n\nMaple did not train V4 Flash. Its product is the privacy and delivery layer around models developed elsewhere. Maple says prompts are encrypted on the user's device, decrypted for inference inside hardware-isolated enclaves and returned through an encrypted connection. Maple also publishes its [client code on GitHub](https://github.com/OpenSecretCloud/Maple?ref=runtimewire) and uses reproducible builds and cryptographic attestation to let users inspect parts of that chain.\n\nThis architecture lets Suman add models as they become available while keeping Maple's central pitch consistent. The product resembles a private model router packaged as a consumer research product, team workspace and OpenAI-compatible developer endpoint.\n\nMaple previously [announced DeepSeek R1 availability](https://blog.trymaple.ai/deepseek-r1-now-available-in-maple-confidential-ai/?ref=runtimewire) for Pro and Team users on June 10, 2025.\n\n### Maple's bet is privacy around open models\n\nMaple lists Pro at $20 a month, Max at $100 a month and Team at $30 per user per month. Its current public materials do not establish V4 Flash-specific usage limits, API rates or additional charges.\n\nThat pricing structure reflects Suman's commercial wager. Maple has to persuade customers that confidentiality is infrastructure worth buying, rather than a setting or contractual promise they expect from a larger AI provider. The strongest prospects are customers handling legal files, financial records, therapy notes, source material or proprietary code, where sending plaintext to an external model operator can create a compliance or trust problem.\n\nMaple says it launched in January 2025 and [reported fivefold user growth](https://blog.trymaple.ai/one-year-of-private-ai-how-maple-ai-made-encryption-easy/?ref=runtimewire) during the six months preceding January 2026, with more than 90% monthly retention among paying customers. Those figures are company-reported and do not include an absolute subscriber count. A [May update](https://blog.trymaple.ai/meet-maple-the-personal-intelligence-platform/?ref=runtimewire) said users had processed millions of chats, thousands of documents and billions of tokens through Maple.\n\nAdding V4 Flash would give those customers another long-context option and give Maple another test of whether model choice can pull users toward its encrypted architecture. It also adds operational work: each model has to be deployed and supported inside Maple's confidential-computing stack.\n\nSuman's advantage is that Maple does not need to win a model-training race. Maple needs to make open models available with a privacy layer customers can understand and verify. V4 Flash would broaden that catalog, but Maple's public materials leave developers without verified Proxy availability, model-specific pricing or usage terms.", "url": "https://wpnews.pro/news/maple-said-it-made-deepseek-v4-flash-available-to-paid-customers", "canonical_source": "https://runtimewire.com/article/maple-deepseek-v4-flash-paid-customers-limited-beta", "published_at": "2026-08-15 01:07:35+00:00", "updated_at": "2026-08-15 01:10:26.539131+00:00", "lang": "en", "topics": ["artificial-intelligence", "ai-products", "ai-infrastructure", "ai-ethics"], "entities": ["Maple", "DeepSeek", "Mark Suman", "Apple", "Instructure", "DeepSeek V4 Flash", "Maple Proxy", "Maple Research"], "alternates": {"html": "https://wpnews.pro/news/maple-said-it-made-deepseek-v4-flash-available-to-paid-customers", "markdown": "https://wpnews.pro/news/maple-said-it-made-deepseek-v4-flash-available-to-paid-customers.md", "text": "https://wpnews.pro/news/maple-said-it-made-deepseek-v4-flash-available-to-paid-customers.txt", "jsonld": "https://wpnews.pro/news/maple-said-it-made-deepseek-v4-flash-available-to-paid-customers.jsonld"}}