Issue 13 · Week 29 of 2026.

Two rows added: kimi-k3 lands Moonshot's 2.8T MoE on the mixture-of-experts branch, recorded as closed until the promised weights and license actually ship, and inkling adds the largest US-origin open-weights model — a 975B trimodal Apache 2.0 MoE from Thinking Machines Lab.

GPT-Red is excluded from the tree by design — OpenAI states it will never be released, and the tree tracks deployed or deployable models. The reported Gemma 4 in-place refresh is weakly sourced (single secondary blog) and is covered under vendor signals rather than as a tree update.

Vendor-stated frontier capability. The releases that reset the closed-source ceiling.

• 2026-07-15/OpenAI/Specialist/Reasoning

  GPT-Red

  An internal-only self-play RL red-teaming model, trained at flagship post-training scale, that beat human red-teamers 84% to 13% on indirect prompt injection — and will never be released

  Procurement should now treat prompt-injection robustness as a quantified, vendor-differentiating spec: GPT-Red hardened GPT-5.6 Sol to a 0.05% direct-injection failure rate (6x fewer than the best model four months prior), and its attack corpora have been folded into every OpenAI release since GPT-5.3 — numbers you can demand equivalents of from every vendor. The withholding itself is the second signal: frontier labs are starting to keep their strongest capabilities as internal advantages, so vendor selection increasingly buys the byproducts of models you will never see. All figures are vendor-reported; the technical preprint promised the week of Jul 20 is the verification checkpoint.

  OpenAI (vendor-reported; preprint promised week of Jul 20); live Andon Labs vending-agent compromise demo

• 2026-07-13/OpenAI via AWS/Frontier/Reasoning

  GPT-5.6 family (Sol / Terra / Luna) on Amazon Bedrock

  Full family GA on Bedrock's Responses API endpoint at OpenAI first-party rates that count toward AWS commitments — with 90% prompt-cache discounts via explicit cache breakpoints

  Teams with AWS enterprise commitments can now route frontier OpenAI traffic through existing spend at no rate premium — re-run the make-vs-route math on any direct OpenAI contract renewal. Mind the caveats before migrating: context is capped at 272K on Bedrock, and Sol is limited to us-east-1/us-east-2 (Terra and Luna add us-west-2), so region-sensitive workloads need the smaller tiers. The 90% prompt-cache discount with explicit breakpoints makes cache design, not list price, the dominant cost lever for agent fleets on this endpoint.

  AWS What's New; Bedrock model card

Open-weights releases that change procurement options. Pull these into pilot when score parity meets license parity.

• 2026-07-16/Moonshot AI/Frontier/MoE

  Kimi K3

  2.8T MoE (16 of 896 experts, ~50B active), 1M context, Kimi Delta Attention, MXFP4 quantization-aware training — hosted-only at launch, with weights promised by Jul 27

  Hold the 'open' celebration until Jul 27: K3 launched API-only, the license is unpublished, the HF repo 404'd the day after launch, and AA classifies it proprietary in the interim — so treat it as a closed model with an open-weights option pending, and gate any self-hosting plan on the actual license text. The capability is real regardless: #1 on the LMArena Frontend Code Arena (first open-weight-lab model to top a flagship board), #4 on the AA Intelligence Index, and the best overall SWE Marathon score (42.0, vendor-reported on the KimiCode harness). Plan for heavy serving if weights do land — Moonshot recommends 64+ accelerator supernodes — and note it launched with only reasoning_effort=max, so token spend runs verbose (130M tokens on the AA eval vs a 63M peer average).

  Moonshot AI; Artificial Analysis; LMArena; Simon Willison

• 2026-07-15/Thinking Machines Lab/Open frontier/MoE

  Inkling

  The largest US-origin open-weights model: 975B/41B-active trimodal MoE under Apache 2.0, 1M context, 45T training tokens, BF16 + NVFP4 checkpoints, day-0 transformers/vLLM/SGLang/llama.cpp

  If your open-model strategy assumed the frontier-scale open lane belonged to Chinese labs, revise it: Inkling is a clean Apache 2.0, US-origin base explicitly positioned for fine-tuning and domain adaptation, with serving partners (Baseten, Modal, Databricks, TogetherAI, Fireworks) live at launch and a managed post-training platform (Tinker) attached. Budget realistically — ~2TB VRAM for BF16 or ~600GB for the calibrated NVFP4 checkpoint — and watch the previewed Inkling-Small (276B/12B active) as the practical deployment tier for most teams. The day-0 NVFP4 release on GB300-trained infrastructure also signals where open-model serving economics are heading: 4-bit datapaths as the default, not an afterthought.

  Thinking Machines Lab via Hugging Face; NVIDIA (GB300 NVL72 training confirmation)

Architectural patterns that crossed multiple vendors this period. Each pattern lists exemplar releases and what it changes for deployment, cost, or capability.

• Extreme MoE sparsity converges on ~2-4% activation at trillion scale

  Kimi K3 (16 of 896 experts, ~1.8%)Inkling (41B of 975B, 4.2%)DeepSeek V4 Pro (49B of 1.6T, ~3%)

  Three independent trillion-scale designs now activate only 2-4% of parameters per token, which means serving cost tracks memory capacity and interconnect bandwidth, not FLOPs — the constraint that matters when sizing inference fleets for these models. It also makes routing stability a first-order engineering concern (Quantile Balancing, Per-Head Muon are the current techniques), because a mis-routed expert at 1.8% activation wastes a much larger share of the useful compute. Capacity planners should price these models on HBM footprint and east-west bandwidth, and treat vendor active-parameter counts as the real sizing number.

  Moonshot AI; Thinking Machines Lab; DeepSeek

• 4-bit-native training collapses the release-to-deployment gap

  Kimi K3 (MXFP4/MXFP8 QAT from the SFT stage)Inkling (calibrated NVFP4 day-0, GB300-trained)

  Both of the week's big releases trained with 4-bit precision in the loop rather than quantizing after the fact — K3 ran quantization-aware training from the SFT stage, and Inkling shipped a calibrated NVFP4 checkpoint the same day as BF16. That removes the accuracy-loss lottery that used to sit between an open-weights release and a production deployment, and it couples open models to FP4-capable datapaths. Procurement implication: hardware refresh decisions should now weight native FP4 support heavily, because the open-model ecosystem is standardizing on it at the source.

  Moonshot AI; Thinking Machines Lab via Hugging Face

• 1M context becomes the default tier — with flat pricing

  Kimi K3 (KDA, flat $3/$15 across 1M)Inkling (1M)DeepSeek Sparse Attention

  Hybrid linear/sparse attention designs (K3's Kimi Delta Attention, DeepSeek Sparse Attention) have made million-token context cheap enough that Moonshot prices it flat — the same $3/$15 per M tokens at 1M as at 1K, with no long-context surcharge. Once long context stops carrying a price penalty, prompt-cache hit rate becomes the dominant cost lever for agent workloads: the difference between a well-designed cache ($0.30 cached-input on K3, 90% discounts on Bedrock GPT-5.6) and a naive one dwarfs the model-choice delta. Architect agent memory around cache breakpoints before optimizing anything else.

  Moonshot AI; AWS Bedrock documentation

Benchmark deltas that change a procurement read. Scores reflect public leaderboards or vendor model cards as of publication.

• LMArena Frontend Code Arena

  Kimi K3 debuted at #1 with 1,679 (preliminary) — the first model from an open-weight lab to top an LMArena flagship board, 48 points clear of Claude Fable 5

  • Kimi K31,679 (prelim) — new #1
  • Claude Fable 51,631
  • GPT-5.6 Sol1,618
  • GLM-5.21,587

  LMArena

• Artificial Analysis Intelligence Index

  K3 entered #4 of 189 at 57.1, above Opus 4.8 — but generated 130M tokens on the eval (63M peer average) at $0.94 per Index task, and AA classifies it proprietary until weights land

  • Claude Fable 559.9 — retains the lead
  • GPT-5.6 Sol58.9
  • Kimi K357.1 — new, #4 of 189
  • Claude Opus 4.856

  Artificial Analysis

• LMArena Text Arena

  K3 entered #9 at 1,486±11 — a 28-place jump over K2.6's #37, moving Moonshot from mid-pack to the top-10 conversation tier in one generation

  • Kimi K31,486±11 — #9, new this week
  • Kimi K2.6#37 (prior generation)

  LMArena

• Agentic coding suite (vendor-reported, KimiCode harness)

  K3 posted the best overall SWE Marathon score and near-frontier terminal results — but every number comes from Moonshot's own harness, and W28's lesson was that harness choice can swing cost and scores 2x, so compare against Sol's Codex-harness numbers with care

  • Kimi K3 — SWE Marathon42.0 (best overall, vendor-reported)
  • Kimi K3 — Terminal-Bench 2.188.3
  • Kimi K3 — FrontierSWE81.2 (Fable 5: 86.6)
  • Kimi K3 — GPQA-Diamond / BrowseComp93.5 / 91.2

  Moonshot AI (vendor-reported, KimiCode harness)

A snapshot of leader-vs-challenger by tier. Useful for procurement shortlists when matching workload to model class. Pair with the benchmark moves above for the underlying scores.

• Closed frontier

  Leader: Claude Fable 5

  Challenger: GPT-5.6 Sol

  Fable 5 keeps the AA Intelligence Index lead (59.9 vs 58.9) and the coding crown; Sol's counter this week is distribution (Bedrock GA at first-party rates) and hardening (0.05% direct-injection failure rate off GPT-Red) rather than raw capability.

• Open frontier

  Leader: GLM-5.2

  Challenger: Kimi K3

  GLM-5.2 holds on the strength of last week's independent Databricks validation; K3 credibly challenges on scores (#4 AA Index, #1 Frontend Code Arena) but stays challenger until the promised weights and license actually ship Jul 27 — Inkling enters the tier as the cleanest-licensed base at this scale.

• Reasoning

  Leader: Claude Fable 5

  Challenger: GPT-5.6 Sol

  Unchanged at the top; K3 launched with only reasoning_effort=max and verbose token behavior (130M tokens on the AA eval), so its reasoning economics are unproven until the promised low/high effort settings arrive.

• Coding

  Leader: Claude Fable 5

  Challenger: Kimi K3

  K3 displaces GLM-5.2 as challenger: #1 Frontend Code Arena (1,679 prelim vs Fable 5's 1,631) and the best SWE Marathon score, though the agentic-suite numbers are vendor-reported on Moonshot's own harness while Fable 5 holds FrontierSWE (86.6 vs 81.2).

• Multimodal

  Leader: Gemini 3.5 Flash

  Challenger: Kimi K3

  Gemini 3.5 Pro missed its reported Jul 17 target — the third slip; K3 enters as challenger with text+image+video input at 1M context, and Inkling's trimodal (text/image/audio) Apache 2.0 base gives builders an open multimodal foundation for the first time at this scale.

• Edge / small

  Leader: Mellum2

  Challenger: Inkling-Small (previewed)

  Inkling-Small (276B/12B active, weights promised after testing) would reset the efficient-deployment tier if it ships as previewed; until then the Hy3 local-quant pipeline from W28 remains the practical near-frontier-at-home path.

The non-release moves that shift vendor risk — pricing, deprecations, gating decisions, license changes — with a one-line procurement read.

• 2026-07-16/Moonshot AI

  K3 pricing resets the Chinese-lab tier: $0.30 cached / $3.00 input / $15.00 output per M tokens — input up 216% and output up 275% versus K2.6, at parity with Claude Sonnet 5

  The 'Chinese open models are 10x cheaper' planning assumption is dead: Moonshot is pricing on capability, not on lane. Re-run any routing or budget model built on that assumption — the K2.x tiers remain available as the value option, but the frontier Chinese tier now costs what the US mid-frontier costs.

  Moonshot AI pricing documentation; Simon Willison

• 2026-07-16/Moonshot AI

  K3 launched as 'open' with no weights: hosted-only, license unpublished until the promised Jul 27 drop, HF repo 404'd the day after launch — AA classifies it proprietary in the interim

  'Open' is becoming a launch-marketing label decoupled from the license-and-weights fact. Contract and compliance teams should gate any K3 self-hosting or fine-tuning commitment on the actual license text, and treat 'weights coming' announcements as vendor roadmap, not product.

  Artificial Analysis; Simon Willison

• 2026-07-14/DeepSeek

  Hard retirement of the deepseek-chat / deepseek-reasoner aliases lands Jul 24, alongside the first time-of-day surge pricing from a major API — 2x listed rates during Beijing peak hours — with V4 GA signaled as imminent

  Surge pricing is a structural first: inference is being priced like a capacity-constrained utility, so cost models that assume flat per-token rates need a time-of-day dimension, and latency-tolerant batch workloads should be scheduled into off-peak windows. The alias retirement is a hard migration deadline — audit for pinned deepseek-chat/deepseek-reasoner IDs now.

  DeepSeek API news

• 2026-07-15/OpenAI

  GPT-Red stays internal: a model trained at flagship post-training scale that OpenAI states will never be released — its output ships only as hardening folded into released models

  This is the clearest capability-gating precedent yet: the strongest capabilities may increasingly arrive as vendor-internal advantages you rent the byproducts of, not products you evaluate. Expect competitors to follow, and adjust diligence accordingly — ask vendors what internal-only systems shaped the model you are buying, and demand the quantified safety deltas (0.05% direct-injection failure is now the number to beat).

  OpenAI (vendor-reported)

• 2026-07-15/Google

  Reported in-place refresh of the Gemma 4 weights, FlashAttention-4 kernels, and chat templates across the HF collection — targeting tool-call JSON consistency, vision OCR, and long-prompt prefill (weakly sourced; single secondary blog)

  If confirmed, in-place weight refreshes under an unchanged model name break reproducibility assumptions: self-hosted Gemma 4 operators should re-pull stale local caches and pin content hashes, not model names, in any evaluation or compliance pipeline. Treat this as a caution flag until Google documents it first-party.

  ExplainX (secondary; not yet confirmed first-party)

Specific model-side catalysts in the next 7–30 days that would change the read materially. Watching these tells us whether the canopy is widening or thinning.

• By Jul 27

  Kimi K3 weights, license, and tech report

  The week's biggest open question: if clean weights and a permissive license land as promised, K3 becomes the strongest open-weights model on record and the open-frontier scorecard flips; if the date slips or the license restricts, the 'open launch without weights' pattern hardens into a playbook.

• Jul 24

  DeepSeek V4 GA + hard alias retirement

  deepseek-chat and deepseek-reasoner go dark the same window V4 is signaled to arrive ('mid-July' per DeepSeek) — the forced migration will produce the first clean V4 adoption data and test whether surge pricing survives contact with customers.

• Week of Jul 20

  GPT-Red technical preprint

  Every GPT-Red number is currently vendor-reported; the preprint is the first chance for independent scrutiny of the 84%-vs-13% red-teaming claim and the Fake Chain-of-Thought injection class, which together justify making injection robustness a procurement spec.

• By Jul 31

  Gemini 3.5 Pro slip watch

  The leaked Jul 17 target passed — the third slip. Prediction-market odds (~81% by Jul 31 on Polymarket) are the only forward signal available and are speculation-grade; treat any capacity or contract planning around a 2M-context flagship as unanchored until Google commits a date.

• No date announced

  Inkling-Small (276B/12B active) weights

  The previewed small tier is the practical deployment target for most teams eyeing the Inkling base — its release (and whether the Apache 2.0 terms carry over) determines whether Thinking Machines' open push reaches beyond 600GB-class serving budgets.