Open-source

DeepSeek V4 Pro and V4 Flash ship under MIT license with 1M-token context — V4 Pro at 1.6T total / 49B active leads Artificial Analysis Index 52 for open weights

Mistral Large 3 and Mistral Small 4 now ship under Apache 2.0 — significant licensing shift from earlier restrictive terms

DeepSeek V4 and the MIT 1M-context frontier — what the most permissive open-weight flagship signals for enterprise procurement

Scale AI revenue doubles to $1.6B annualized — Meta's data-labeling and RLHF spend the dominant driver, IPO timing accelerated to 2027

EleutherAI releases SAE-Bench 2 — public benchmarking suite for sparse-autoencoder interpretability adds Claude Opus 4.7, GPT-5.2, and Gemini 3.1 Ultra evaluation tasks

Meta releases Llama 4.1 70B Thinking — open-weight reasoning model lands with day-zero llama.cpp and vLLM support, MIT-aligned community license updated

Mistral Large 3 launches under Apache 2.0 — 200B mixture-of-experts model ships with reasoning-mode variant, commercial-deployment terms unchanged from prior open releases

Llama 4.1 and Mistral Large 3 — when the open-thinking-tier finally arrives at parity with the closed frontier

DeepMind releases Gemma Scope 2 as the largest open-source mechanistic interpretability toolkit — sparse-autoencoder features at frontier-scale coverage

Cohere acquires Aleph Alpha — European AI consolidation accelerates as transatlantic capability landscape reshapes

Apache 2.0 holds as open-weight licensing default through mid-2026 — Qwen 3.6 Max-Preview closed pivot remains strategic outlier

Cohere–Aleph Alpha and European consolidation — when the capital ceiling reshape forces non-US lab combination

Apache 2.0 wins the permissive-license race for open-weight LLMs — enterprise procurement friction drops to near-zero through mid-2026

Closed-weight bifurcation at flagship tier — Qwen 3.6 Max-Preview's April pivot tested whether the open-weight default holds at very largest models

Apache license discipline after the Qwen pivot — when the open-weight default holds even as flagship strategies bifurcate

Alibaba's Qwen 3.6 Max-Preview goes CLOSED-WEIGHT on April 20 — first proprietary flagship in Qwen history reshapes open-weight discipline question

Five frontier-class open-weight LLMs shipped in last 30 days — Llama 4, Qwen 3.5, DeepSeek V4, Gemma 4, Mistral Medium 3.5, Apache 2.0 dominates

Qwen 3.6 Max-Preview goes closed-weight and the open-weight discipline gets tested — Alibaba's strategic outlier or the start of a pattern

Alibaba ships Qwen 3.7 Max on May 20 — Chinese frontier-tier coding and reasoning model lands in same week as Google I/O

DeepSeek V4-Pro lands at 1.6T total / 49B active under MIT license — V4-Flash at 284B/13B active for cost-efficient frontier-tier reasoning

Qwen 3.6 Max-Preview holds #1 on six coding and agent benchmarks simultaneously — SWE-bench Pro, Terminal-Bench 2.0, SkillsBench among them

DeepSeek V4 and MIT as strategic weapon — when the open-weight frontier is also the most permissive license

DeepMind ships Gemma Scope 2 and Anthropic open-sources its circuit tracer — mechanistic interpretability infrastructure becomes accessible to independent researchers

Five frontier-class open-weight LLMs shipped in last 30 days — Llama 4, Qwen 3.5, DeepSeek V4, Gemma 4, Mistral Medium 3.5

Moonshot Kimi K2.6 ships at 1T parameters with 32B active, 256K context, and native image and video understanding — April 20 release reframes Chinese open-weight ceiling

Open-weight frontier as default — five families in 30 days and the procurement math flips

Mistral closes $2B Series E — Bpifrance leads, $14B post-money brings European-flagship valuation gap with US frontier tier to 27x

Anthropic opens Circuits public eval — releases activation-probe dataset for community red-team of interpretability claims

Meta ships Llama 5.1 405B — open-weights flagship returns to monthly cadence, reasoning-tier eval gaps with closed-flagships narrow further

DeepSeek R2 launches — China's open-weights reasoning frontier matches GPT-5 on AIME, undercuts inference cost by 11x

Open-weights after the Llama pause — Llama 5.1 and DeepSeek R2 reset the trajectory the closed frontier had been pulling ahead of

Deep Cogito v2: open-source models that internalize their own reasoning

San Francisco startup founded by ex-Googlers ships four open-source hybrid reasoning models — 70B, 109B, 405B, 671B — using a technique called Iterated Distillation and Amplification (IDA) to distill search-time reasoning back into model weights.

Microsoft Phi-4 family expands: -mini, -multimodal, -reasoning, -reasoning-vision

Microsoft's small-language-model bet now includes Phi-4-mini, Phi-4-multimodal (text+audio+vision in one), Phi-4-reasoning, Phi-4-reasoning-plus, Phi-4-mini-reasoning, and Phi-4-reasoning-vision. Reportedly beats DeepSeek-R1-Distill-Llama-70B at most benchmarks despite far smaller size.

DeepSeek V4-Flash holds 1M context under MIT — 284B/13B-active MoE proves the Flash-tier-open-frontier convergence

DeepSeek's V4-Flash variant (284B total / 13B active parameters, 1M context, MIT license) holds production-tier capability at hyperscaler-routable scale. Combined with V4-Pro (1.6T total / 49B active, 80.6 SWE-Bench Verified, 90.1 GPQA Diamond), DeepSeek now ships the most operationally credible open-weight Pro/Flash split. The 1M context retention in Flash is the structural detail that erases the case for routing to Pro on long-document workloads.

DeepSeek V4 Pro vs Flash — the procurement decision tree clarifies at MIT-licensed weights

DeepSeek's V4 release (April 24) shipped two SKUs: V4-Pro (1.6T total / 49B active parameters, 80.6 SWE-Bench Verified, 90.1 GPQA Diamond) and V4-Flash (284B total / 13B active, 1M context). Both run under the MIT license, both ship at 1M context, and both clear the bar for production deployment on coding and reasoning workloads. The Pro/Flash bifurcation now mirrors the closed-flagship pricing curve at a fraction of the cost.

Mistral Medium 3.5 lands as the EU-friendly coding pick — 77.6% SWE-Bench at sovereign-jurisdiction licensing

Mistral Medium 3.5 (April 29 release) lands at 77.6% on SWE-Bench Verified with EU-friendly licensing terms — the strongest sovereign-jurisdiction coding-model offering in the May 2026 lineup. Combined with Mistral Large 3 (675B / 41B active MoE) and the Voxtral TTS, Forge, and Leanstral releases earlier in the year, Mistral's 2026 H1 cadence is closer to Qwen's monthly tempo than to its prior quarterly pattern.

Qwen 3.6-35B-A3B and Qwen 3.6-27B ship as open weights — Alibaba presses the cadence advantage with monthly drops

Alibaba's Qwen 3.6-35B-A3B (Apr 2026) and Qwen 3.6-27B (Apr 2026) continue the team's roughly-monthly drop cadence across 2026 H1. Combined with Qwen 3.5 (Feb 2026, 397B MoE with unified vision-language and 201 languages) and Qwen 3.6 Plus / Max Preview (Apr 2/20), Alibaba now ships the most operationally aggressive open-weights release schedule among Tier 1 labs.

The open-weight frontier is now constraint-routable — Qwen for cadence, DeepSeek for MIT, Mistral for EU

Three labs occupy the open-weight Tier 1 ladder. Each serves a different procurement constraint. The 'open-weight model selection' decision has stopped being a single comparison and become a constraint-mapping exercise. That's a healthier market than the one we had six months ago.

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.

Chinese open-weight models now account for more than 60% of OpenRouter usage — a 60× jump in 18 months

Air Street's State of AI May 2026 report shows Chinese open-weight models — DeepSeek, Qwen, Kimi, GLM — went from roughly 1% of OpenRouter usage in mid-2024 to more than 60% in May 2026. The shift tracks a 5–20× price-per-token gap to closed flagships and a near-elimination of the capability gap on most evaluation suites.

Deep Cogito v2 ships 70B/109B/405B/671B open-weight family with Iterated Distillation & Amplification self-improvement loop

Deep Cogito's v2 release ships four open-weight sizes (70B, 109B, 405B, 671B) wired into an Iterated Distillation & Amplification (IDA) self-improvement loop. The release positions IDA as a deployable architecture rather than a research curiosity — the first open-weight family where the "model improves itself between checkpoints" methodology is shipped as the default training recipe.

DeepSeek V4 Flash quietly extends 1M context to standard tier — Apache-2.0 weights match closed-flagship reasoning on Pass@1

DeepSeek extended the 1M context window to its V4 Flash tier this week, pushing the cheaper standard SKU into a capability bracket previously occupied only by V4 Pro and closed flagships. Combined with the unchanged 80.6% SWE-Bench Verified ceiling and the MIT/Apache-2.0 license, the practical effect is to compress the price-quality gradient on long-context production workloads.

Mistral Medium 3.5 ships as the EU-friendly coding pick — 77.6% SWE-Bench Verified at open-weight Apache pricing

Mistral Medium 3.5, released April 29 and now widely available across cloud providers, hit 77.6% SWE-Bench Verified — putting it within striking distance of Qwen 3.5 and DeepSeek V4 on coding while shipping under Apache 2.0 from a Paris-based lab. For EU enterprises navigating data-residency-plus-IP-clarity procurement constraints, the model is the most defensible production-tier coding choice currently available.

The China-share tipping point — when did the OpenRouter graph cross 50%?

Sometime in early 2026, Chinese open-weight models crossed 50% of OpenRouter usage. The exact moment matters less than the realization: production share has already migrated. The policy conversation is debating a battle that's already moved one front forward.

The Flash tier becomes the frontier — capability per dollar, not capability per parameter

Gemini 3.5 Flash hits 76.2% Terminal-Bench. DeepSeek V4 Flash gets 1M context. Mistral Medium 3.5 hits 77.6% SWE-bench Verified at Apache pricing. The 2026 frontier isn't the highest-capability model — it's the highest-capability-at-Flash-pricing model.

DeepSeek V4 ships under MIT license — 1.6T Pro and 284B Flash, both at 1M context

DeepSeek released V4 (Pro at 1.6T total / 49B active, Flash at 284B total / 13B active) on April 24 under MIT licensing. Both variants ship with 1M token context. V4 Flash pricing of $0.14/M input is the floor for the open-weight frontier and is forcing competing labs to reprice or differentiate on capability.

Kimi K2.6, GLM-5.1, and Qwen 3.6 27B ship in the same fortnight — open-weight pace doubles

Three open-weight releases in two weeks: Moonshot Kimi K2.6 (top-tier coding, 1T total / 32B active, 256K context), Z.ai GLM-5.1 ($0.18/M input), and Qwen 3.6 27B (77.2% SWE-bench Verified). The open-weight pace has now compressed to roughly one Pro-tier release per week.

Meta Llama 4 and Mistral Medium 3.5 anchor the European-American open-weight tier

Meta shipped Llama 4 in April 2026 with Scout (17B active / 109B total MoE, runnable on 10GB VRAM) and Maverick (17B active / 400B total). Mistral Medium 3.5 launched April 29 — a 128B dense model hitting 77.6% on SWE-bench Verified, the best single-vendor coding stack outside the Anthropic and OpenAI labs.

Mistral Large 3 ships as 675B / 41B sparse MoE under Apache 2.0

Mistral Large 3 lands as a 675B-total / 41B-active sparse Mixture-of-Experts model under Apache 2.0 licensing. The architecture choice mirrors DeepSeek V4 and Llama 4 Maverick — the open-weight tier has converged on sparse MoE as the default frontier architecture.

The open-weights rebound: capability parity at one-tenth the price

DeepSeek V4 under MIT, GLM-5.1 at $0.18/M, Kimi K2.6 at 256K context, Llama 4 Maverick. The open-weight frontier is now within a few SWE-bench points of closed flagships at one-tenth the input cost. The structural implications run deeper than pricing.

Four Chinese open-weights labs shipped frontier-class models in a 12-day window

Z.ai (GLM-5.1), MiniMax (M2.7), Moonshot (Kimi K2.6), and DeepSeek (V4) all landed in a 12-day window in early-to-mid May 2026 — all clearing 75%+ on SWE-bench Verified, all priced below $0.30/M input tokens, all permissively licensed for commercial use.

Meta confirms open-source Avocado and Mango variants alongside closed flagships

Meta has confirmed it will release open-weights versions of its next two frontier models, codenamed Avocado and Mango, while keeping the largest variants proprietary — a hybrid strategy that splits the difference between Llama's open-source heritage and the closed-model economics of rival labs.

Z.ai GLM-5.1 ships at $0.18 per million input tokens with 1M context

Z.ai released GLM-5.1 with 1 million token context and inference pricing of $0.18 per million input tokens — undercutting DeepSeek V4 Flash ($0.14) only narrowly while matching it on SWE-bench Verified at 76.4%.

DeepSeek V4 Pro: 80.6 SWE-Bench Verified, 1M context, sub-$0.20 per million tokens

DeepSeek shipped V4 Pro (and V4 Flash) on Hugging Face and the official API. Headline numbers: 80.6 SWE-Bench Verified, 90.1 GPQA Diamond, 1M token context. V4 Flash undercuts most frontier pricing at $0.14 per million input tokens.

Qwen 3.6 27B hits 77.2% SWE-bench while staying small enough for single-GPU inference

Alibaba released Qwen 3.6 27B with a 77.2% SWE-bench Verified score — a frontier-competitive number on a model small enough to run on a single H100. The 27B parameter sweet spot has become the most-shipped open-weights size of 2026.

Moonshot releases Kimi K2.6 — current state-of-the-art for open-weights coding

Moonshot AI shipped Kimi K2.6, a coding-specialized open-weights model that posts the strongest SWE-bench Verified score among open releases — narrowly ahead of DeepSeek V4 Pro on multi-file edits.

IBM ships Granite 4.1: dense 8B that matches 32B MoE, plus vision/speech/safety variants — all Apache 2.0

Granite 4.1 covers 3B / 8B / 30B language models, Granite Vision 4.1 (top score on 7 chart/table/KVP extraction benchmarks), two ASR speech models, embeddings, and a Granite Guardian 4.1 safety classifier — every variant under Apache 2.0. The 8B dense model reportedly matches or beats 32B MoE systems.

Mistral Medium 3.5 lands — capstone on a six-week release blitz

Mistral Medium 3.5 (Apr 29) is a frontier multimodal model targeted at agentic and coding workloads. It's the headline at the end of a stretch where Mistral shipped Small 4 (unifying Magistral/Pixtral/Devstral), Voxtral TTS, Leanstral for formal proofs, and the Forge enterprise platform — all between March 16 and end of April.

Mistral Medium 3.5 — 128B dense, 256K context, 77.6% on SWE-Bench Verified

Mistral's April 29 release ships under a modified MIT license, with 77.6% on SWE-Bench Verified — positioning the model ahead of Devstral 2 and Qwen 3.5 397B A17B at a fraction of the active-parameter budget.

NVIDIA ships Nemotron 3 Nano Omni — 30B hybrid Mamba-Transformer MoE (3B active), multimodal for agents

Nemotron 3 Nano Omni (April 28) unifies vision, audio, language, and text into one open multimodal model. The architecture is the interesting bit: a hybrid Mamba-Transformer MoE with 30B parameters and only 3B activated per forward pass.

NVIDIA Nemotron 3 Super — 120B hybrid MoE (12B active) tuned for local agent deployment

NVIDIA's open Nemotron 3 Super lands as a 120B-parameter hybrid MoE with 12B active and a 1M-token context window. The explicit design target: local agent deployment with tool-augmented coding workloads.

NVIDIA Nemotron 3 Nano Omni — unified vision, audio, language for agents

NVIDIA's open Nemotron 3 Nano Omni unifies vision, audio, and language processing in a single model, claiming up to 9x efficiency improvement for agent workloads versus equivalent stacks of specialist models.

Alibaba ships Qwen 3.6 Plus and Qwen 3.6 Max Preview — agentic push in two-week tempo

Qwen 3.6 Plus dropped April 2; Qwen 3.6 Max Preview followed April 20. Alibaba's framing: "accelerating agentic AI deployment for enterprises and Alibaba's AI applications." Built on the Qwen 3.5 native-multimodal foundation from February, which supports 201 languages.

Google Gemma 4 ships under Apache 2.0 — four sizes, MoE, multimodal, 256K context

Gemma 4 (April 2) arrives in E2B / E4B / 26B MoE / 31B Dense variants with native image+video everywhere and native audio on the smaller models. 256K context, 140+ languages, agentic-workflow-oriented. The 31B Dense reportedly hit #3 on Arena's text leaderboard.

Alibaba Qwen 3.5 Omni — native multimodal text/audio/video with sub-300ms TTFT

Qwen 3.5 Omni (released March 30) is a native multimodal model handling text, audio, video, and real-time interaction. Real-time audio time-to-first-token comes in below 300ms with 95%+ ASR accuracy — the relevant numbers for actual voice-assistant deployment.