OpenMed Router Blog
HIPAA-compliant AI for healthcare
Practical guides on compliance, open source model selection, and running the latest models under a BAA. Written by Chris Williams, MD.
Model launches
New open source models, their specs, and what they mean for HIPAA-compliant healthcare inference.
September 1, 2026
Introducing HIPAA-compliant DeepSeek-V4-Flash-Vision-Exp
Flash-Vision-Exp is the first multimodal DeepSeek: text-plus-image input at Flash-class prices, with agentic scores barely moved from the text-only checkpoint.
Read more →August 29, 2026
Introducing HIPAA-compliant GLM-5.3
GLM-5.3 keeps GLM-5.2's 753B MoE base and gets all its gains from a month of aggressive post-training, with Terminal-Bench 3.0 the first public benchmark where an open model led the frontier.
Read more →August 27, 2026
Introducing HIPAA-compliant GLM-5.3-Flash
GLM-5.3-Flash packs first-multimodal GLM-5 into 320B total and 18B active parameters under MIT, with a hybrid attention design and a 1M-token context window.
Read more →August 27, 2026
Introducing Qwen3.8-Flash-Next: Qwen4's architecture
Flash-Next is a technical preview of the Qwen4 architecture, not a finished product: 125B MoE at 6B active with n-gram embeddings and a new sparse attention design.
Read more →August 15, 2026
Introducing HIPAA-compliant Qwen3.8-27B
Qwen3.8-27B packs native vision, agentic coding near Opus-class benchmarks, and an Apache 2.0 license into a dense 27B model that runs on a GPU you already own.
Read more →August 14, 2026
Introducing HIPAA-compliant DeepSeek-V4-Pro-0813
DeepSeek-V4-Pro-0813 is the general availability release of DeepSeek's 1.6T flagship, with the biggest agentic gains of any checkpoint this year.
Read more →August 14, 2026
Introducing Qwen3.8-Max open weights
Qwen3.8-Max's open weights are the largest open-weight release ever at 2.4T parameters. Here is what is actually inside, and what it takes to run.
Read more →August 3, 2026
Introducing HIPAA-compliant DeepSeek-V4-Flash-0731: the flash model that outruns its Pro sibling
DeepSeek-V4-Flash-0731 beats DeepSeek-V4-Pro (Preview) on nearly every agentic and coding benchmark - DeepSWE 54.4, Cybergym 76.7, Terminal Bench 82.7 - at $0.14/M input, HIPAA-compliant under a BAA.
Read more →July 30, 2026
Introducing HIPAA-compliant Kimi K3: frontier-class intelligence for healthcare
Kimi K3, the #1 open-weight model on the Artificial Analysis Intelligence Index (score 57), is now available HIPAA-compliant under a BAA - 1M-token context, native vision, and frontier reasoning at open-weights pricing.
Read more →July 5, 2026
Introducing HIPAA-compliant GLM 5.2: frontier-class reasoning for healthcare
GLM 5.2 from Z.ai is the highest-scoring open-weight model on the Artificial Analysis Index v4.1. Here's what it's good at, where it falls short, and how to run it under HIPAA.
Read more →July 4, 2026
Introducing HIPAA-compliant DeepSeek V4 Pro: flagship-tier intelligence for high-stakes healthcare AI
DeepSeek's 1.6T flagship scores 90.1% on GPQA Diamond and 80.6% on SWE-bench Verified. Here's what it's good at, where it falls short, and how to run it under HIPAA.
Read more →July 3, 2026
Introducing HIPAA-compliant DeepSeek V4 Flash: top-tier speed at a fraction of the cost
DeepSeek-V4-Flash runs at 200+ tokens/second for $0.14/M input. Here's what it's good at, where it falls short, and how to run it under HIPAA.
Read more →
Provider comparisons
How OpenAI, Anthropic, Azure, and Bedrock stack up against open source inference for healthcare workloads.
July 10, 2026
Anthropic Claude for healthcare: BAAs, pricing realities, and the open-source alternative
Claude Opus 4.8 costs $25/M output tokens; Haiku 4.5 costs $5/M. GLM 5.2 and DeepSeek V4 Flash deliver comparable tiers through a BAA-backed pipeline at up to 18x lower cost.
Read more →July 10, 2026
OpenAI for healthcare: understanding BAAs, compliance pipelines, and the 10x open-source price advantage
OpenAI's GPT-5.6 Sol costs $30/M output tokens. GLM 5.2 and DeepSeek V4 Pro deliver frontier-class reasoning through a BAA-backed pipeline at a fraction of the cost. Here's the full comparison.
Read more →
Compliance
HIPAA, BAAs, and what compliant AI inference actually requires for healthcare teams.
Foundations
Plain-English definitions for the AI concepts health-tech teams need to know.
July 1, 2026
How we evaluate AI models for healthcare (our editorial standards)
How we assess open source models, verify compliance claims, and keep health-tech teams up-to-date with accurate information.
Read more →July 1, 2026
What does 'frontier AI' actually mean? A 2026 definition for healthcare teams
Frontier AI has moved from research labs into executive orders and compliance frameworks. Here's what it means, which models hold the title in 2026, and why it matters for health-tech teams.
Read more →