The monthly drop cadence — Qwen's release schedule is becoming the open-weight default tempo

Qwen 3.5 in February. Qwen 3.6 Plus in April. Qwen 3.6 Max Preview also April. Qwen 3.6-35B-A3B and 3.6-27B as open weights. Five major releases in twelve weeks. Mistral and Meta ship slower; Alibaba is teaching the rest of the open-weight community what monthly cadence looks like.

The release schedule data

Qwen 3.6-35B-A3B and Qwen 3.6-27B dropped as open weights in April. Combined with Qwen 3.5 (February), Qwen 3.6 Plus (April 2), and Qwen 3.6 Max Preview (April 20), Alibaba ships roughly one major release per month. Mistral ships quarterly. Meta's Llama 5 has been in development since the Llama 4 launch in April 2025.

Why cadence is a competitive variable

The 2024 mental model — that open-weight releases happen on an annual cycle, with major bumps tied to compute build-outs and tokenizer regenerations — is broken. The 2026 reality is that open-weight Tier 1 labs ship incremental capability improvements on monthly cycles, with the team that ships fastest establishing the latest procurement comparable.

For procurement teams, the implication is operational. Pinning an open-weight model to a single version (Llama 4 Maverick, for example) means accepting that the procurement comp ages within weeks. Either set up automated re-evaluation against the latest Qwen monthly drop, or accept that your routing layer is running on benchmarks from two months ago.

Monthly drops compound. Annual releases don't.

The complementary trend: Flash tiers

DeepSeek V4 Pro and V4 Flash follow a parallel architectural move: ship cheap-tier and frontier-tier weights together, both under MIT/Apache licensing, both at production capability. The cadence-plus-Flash combination is the open-weight ecosystem's structural advantage over the closed-flagship tier: faster iteration, broader access, better per-dollar capability.

The forward read

The 2027 open-weight landscape will likely look like: 2-3 dominant families (Qwen, DeepSeek, plus one US-based open team), each shipping monthly drops, each with a Flash/Pro split, each under permissive licensing. The Meta Llama franchise either matches the cadence or accepts a competitive position behind it. Mistral has the closest competitive posture but slower cadence; their May 2026 update tempo will reveal whether they can match Qwen's pace.

llm-stats — LLM model updates → · Codersera — open-source LLM landscape 2026 → · HuggingFace — best open-source LLMs →