---
title: >-
  Washington entered the release pipeline: GPT-5.6 launched gated to ~20 approved partners, Fable 5
  came back, and Sonnet 5 split sticker price from cost-per-task.
publication: The Model Pulse
slug: 2026-W27
issueNumber: 11
isoYear: 2026
isoWeek: 27
cadence: weekly
publishedAt: '2026-07-04'
periodLabel: Week 27 of 2026
canonicalUrl: https://brianletort.ai/industry/models/2026-W27
pdfUrl: https://brianletort.ai/downloads/model-pulse-2026-W27.pdf
schemaVersion: 2026.05.02
treeDelta:
  added:
    - claude-sonnet-5
    - gpt-5-6-sol
    - longcat-2-0
    - leanstral-1-5
    - nemotron-labs-twotower
  updated:
    - claude-fable-5
frontierMovements:
  - modelId: claude-sonnet-5
    name: Claude Sonnet 5
    vendor: Anthropic
    releaseDate: '2026-06-30'
    tier: reasoning
    architecture: agentic
  - modelId: gpt-5-6-sol
    name: GPT-5.6 family (Sol / Terra / Luna)
    vendor: OpenAI
    releaseDate: '2026-06-26'
    tier: specialist
    architecture: reasoning
  - modelId: claude-fable-5
    name: Claude Fable 5
    vendor: Anthropic
    releaseDate: '2026-07-01'
    tier: frontier
    architecture: reasoning
openWeights:
  - modelId: longcat-2-0
    name: LongCat-2.0
    vendor: Meituan
    releaseDate: '2026-06-30'
    tier: open_frontier
    architecture: moe
  - modelId: leanstral-1-5
    name: Leanstral 1.5
    vendor: Mistral AI
    releaseDate: '2026-06-30'
    tier: specialist
    architecture: moe
  - modelId: nemotron-labs-twotower
    name: Nemotron-Labs-TwoTower
    vendor: NVIDIA
    releaseDate: '2026-07-01'
    tier: specialist
    architecture: moe
architecturePatterns:
  - Effort dials decouple sticker price from cost-per-task
  - Diffusion decoding retrofitted onto frozen AR backbones
  - 1M-context sparse attention is the new open-weight table stakes
  - Training and serving substrates diversify away from NVIDIA
benchmarks:
  - Artificial Analysis Intelligence Index (v4.1)
  - SWE-bench Pro (vendor-scaffold aggregate)
  - LMArena
  - PutnamBench / miniF2F (formal math)
scorecardAsOf: '2026-07-04'
vendorSignals:
  - vendor: OpenAI + Anthropic (US Commerce)
    date: '2026-06-26'
    signal: >-
      Within five days the US government gated a new frontier launch (GPT-5.6, ~20 approved
      partners) and un-gated a suspended one (Fable 5 restored July 1)
  - vendor: Anthropic
    date: '2026-06-30'
    signal: >-
      Sonnet 5 launched at introductory $2/$10 per M tokens through August 31 (then $3/$15); Fable 5
      moves to usage credits even on paid plans after July 7
  - vendor: DeepSeek
    date: '2026-07-24'
    signal: >-
      Legacy deepseek-chat / deepseek-reasoner aliases go fully dark July 24 15:59 UTC; they
      currently silently route to deepseek-v4-flash
  - vendor: Google
    date: '2026-06-30'
    signal: >-
      Veo 2.0/3.0 video models shut down June 30; same day, Nano Banana 2 Lite hit GA and Gemini
      Omni Flash entered public preview at $0.10/second of 720p video
  - vendor: xAI
    date: '2026-07-01'
    signal: >-
      Voice Agent Builder beta: no-code speech-to-speech agents on Grok Voice at $0.05/min audio +
      $0.01/min telephony — roughly 1/3 to 1/5 of ElevenLabs/Vapi-class pricing
  - vendor: Anthropic
    date: '2026-06-30'
    signal: >-
      Claude Science beta shipped — a multi-agent scientific workbench on existing Claude models,
      with up to 50 funded projects at $30K credits each (applications through July 15)
---

# Washington entered the release pipeline: GPT-5.6 launched gated to ~20 approved partners, Fable 5 came back, and Sonnet 5 split sticker price from cost-per-task.

*The Model Pulse · Issue 11 · Week 27 of 2026 · Published 2026-07-04*

## Big Read

Within five days the US government both gated a new frontier launch and un-gated a suspended one. OpenAI previewed the GPT-5.6 family (Sol, Terra, Luna) on June 26 to roughly 20 government-approved partner organizations — the first US frontier model launched under a government-managed access list — while Commerce withdrew the Anthropic export-control order on June 30 and Claude Fable 5 returned globally on July 1. The procurement implication is direct: frontier availability is now partly a regulatory variable, so contracts and architectures must assume any flagship can be gated at launch or suspended after it, and multi-model routing is availability insurance, not just cost optimization. The week's biggest GA release, Claude Sonnet 5 (June 30, 1M context, 85.2% SWE-bench Verified, default for Claude Free/Pro), delivered the second structural lesson: at max effort its token appetite pushes measured cost to $2.29 per task — ~15% above Opus 4.8 — despite a far lower per-token price, so sticker price and cost-per-task have formally diverged. Buyers should route by measured cost-per-completed-task with effort level as an explicit parameter, and re-run that math before Sonnet 5's introductory pricing lapses on August 31. Meituan's MIT-licensed LongCat-2.0 rounds out the week by adding demand-proven open-weight pressure — and a claimed NVIDIA-free trillion-scale training run — to the cost lane.

## Tree delta

Five rows added: claude-sonnet-5 and gpt-5-6-sol extend the closed reasoning canopy, longcat-2-0 lands a 1.6T open MoE, leanstral-1-5 opens a formal-verification specialist node, and nemotron-labs-twotower extends the diffusion-retrofit lineage. One row updated: claude-fable-5 now records the July 1 restoration.

**Added (5):** `claude-sonnet-5`, `gpt-5-6-sol`, `longcat-2-0`, `leanstral-1-5`, `nemotron-labs-twotower`.

**Updated (1):** `claude-fable-5`.

*gpt-5-6-sol carries status gated (preview limited to ~20 government-approved partners, no independent benchmarks). The unverified Pulpie family (Feyn Inc.) was excluded as below the significance and sourcing threshold.*

## Frontier movements

### Claude Sonnet 5 `claude-sonnet-5`

**Anthropic** · 2026-06-30 · reasoning · agentic

_The week's biggest GA release: 1M context, 85.2% SWE-bench Verified, effort dials, at introductory $2/$10 per M tokens_

Sonnet 5 gives architects near-Opus agentic quality at a mid-tier sticker price — independent testing shows it beating Opus 4.8 on agentic knowledge-work benchmarks — but at max effort it costs ~15% more per completed task than Opus 4.8. Route by measured cost-per-task, not per-token price, and model post-promo economics (standard $3/$15 from September 1) before committing volume.

Source: [Anthropic; Artificial Analysis](https://www.anthropic.com/news/claude-sonnet-5).

### GPT-5.6 family (Sol / Terra / Luna) `gpt-5-6-sol`

**OpenAI** · 2026-06-26 · specialist · reasoning

_Next OpenAI flagship generation previewed — but launched to only ~20 government-approved partner organizations_

Marked specialist because it is gated: this is the first US frontier launch under a government-managed access list, so procurement teams should treat release gating as a live availability risk for every future frontier upgrade and keep multi-vendor fallbacks tested. Terra's claimed GPT-5.5-level capability at half the price is the pricing lever to model ahead of GA.

Source: [OpenAI](https://openai.com/index/previewing-gpt-5-6-sol/).

### Claude Fable 5 `claude-fable-5`

**Anthropic** · 2026-07-01 · frontier · reasoning

_Restored globally July 1 after Commerce withdrew the June 12 export-control order; back atop the available boards_

The strongest model on the public leaderboards is purchasable again, resetting the practical-leader question W26 settled in Opus 4.8's favor — but the 19-day forced outage is now a demonstrated failure mode, so any single-model dependency on a frontier flagship needs a tested fallback path. Note the fine print: after July 7, Fable 5 moves to usage credits even on paid plans, so treat it as a premium metered resource in budgets.

Source: [Anthropic; VentureBeat](https://www.anthropic.com/news/redeploying-fable-5).

## Open weights

### LongCat-2.0 `longcat-2-0`

**Meituan** · 2026-06-30 · open_frontier · moe

_OpenRouter's stealth chart-leader 'Owl Alpha' unmasked: a 1.6T MoE with native 1M context under MIT, trained on 50K+ domestic Chinese ASICs_

A near-frontier agentic-coding MoE with two months of real developer demand joins GLM-5.2 and DeepSeek V4 in the open cost-pressure lane — benchmark it wherever unit economics beat absolute frontier quality, but verify actual weight availability before treating it as self-hostable. The claimed NVIDIA-free 35T-token training run, if it holds, weakens the assumption that export controls throttle Chinese frontier training; factor that into sovereignty and supply planning.

Source: [Hugging Face model card; VentureBeat; SiliconANGLE](https://huggingface.co/meituan-longcat/LongCat-2.0).

### Leanstral 1.5 `leanstral-1-5`

**Mistral AI** · 2026-06-30 · specialist · moe

_Apache 2.0 Lean 4 proof-engineering agent (119B MoE / 6.5B active) saturates miniF2F and solves 587/672 PutnamBench problems at ~$4 each_

The strongest evidence yet that narrow verticalized agent models can beat frontier brute force on economics — ~75x cheaper per problem than Seed-Prover 1.5 high — with a practical hook for engineering leaders: Mistral reports 5 previously unknown bugs found across 57 open-source repositories via formal code verification. Self-host the Apache weights for anything durable; the free labs API endpoint retires September 30.

Source: [Mistral AI; MarkTechPost](https://mistral.ai/news/leanstral-1-5/).

### Nemotron-Labs-TwoTower `nemotron-labs-twotower`

**NVIDIA** · 2026-07-01 · specialist · moe

_Open block-diffusion model on a frozen AR backbone: 2.42x wall-clock throughput at 98.7% of autoregressive quality_

TwoTower shows diffusion decoding can be retrofitted onto an existing pretrained AR checkpoint for ~2.1T training tokens instead of a 25T re-pretrain — a cheap serving-throughput upgrade path any lab with a strong AR model can copy, which compounds W26's serving-economics story. One checkpoint supports diffusion, mock-AR, and standard AR decoding, so platform teams can A/B the pattern with low switching risk once vLLM-class support lands.

Source: [arXiv 2606.26493; MarkTechPost](https://arxiv.org/abs/2606.26493).

## Architecture watch

### Effort dials decouple sticker price from cost-per-task

_Examples:_ Claude Sonnet 5, DeepSeek V4, GPT-5.5.

Every major vendor now ships user-controllable reasoning-effort settings, and this week produced the clearest data point yet that token-hungry effort scaling inverts price intuition: Sonnet 5 at max effort costs $2.29 per Intelligence Index task at standard pricing — more per completed task than the nominally pricier Opus 4.8. Evaluation harnesses must price the task, not the token, and effort level must become an explicit routing parameter in every model gateway.

Source: [Artificial Analysis](https://artificialanalysis.ai/articles/claude-sonnet-5-agentic-cost).

### Diffusion decoding retrofitted onto frozen AR backbones

_Examples:_ Nemotron-Labs-TwoTower, DiffusionGemma.

Two vendors in a month have shipped open diffusion LLMs built on top of existing autoregressive checkpoints rather than trained from scratch — TwoTower needed only ~2.1T training tokens against its backbone's 25T. Diffusion is emerging as a cheap post-hoc throughput upgrade (2.42x wall-clock at 98.7% quality) rather than a rival pretraining paradigm; platform teams should watch for this reaching production serving stacks via vLLM-class support.

Source: [arXiv 2606.26493](https://arxiv.org/html/2606.26493v2).

### 1M-context sparse attention is the new open-weight table stakes

_Examples:_ DeepSeek V4, LongCat-2.0, GLM-5.2.

The Chinese open-weight labs have converged on the same recipe: sparse/compressed attention variants that make 1M-token context economically routine. The DeepSeek V4 technical report published this window puts hard numbers on it — roughly 90% KV-cache reduction and 27% of V3.2's inference FLOPs at 1M tokens — which directly attacks the HBM-scarcity constraint W26 flagged. Long-context agentic workloads are becoming an open-weight strength, not a closed-model premium, so re-benchmark long-context routing assumptions.

Source: [DeepSeek V4 technical report (arXiv 2606.19348)](https://arxiv.org/html/2606.19348).

### Training and serving substrates diversify away from NVIDIA

_Examples:_ LongCat-2.0, GPT-5.6 Sol on Cerebras, Groq inference cloud.

In one week, a claimed frontier-scale non-NVIDIA training run (LongCat-2.0 on 50K+ domestic Chinese ASICs) and a flagship non-NVIDIA serving deal (GPT-5.6 Sol on Cerebras at up to 750 tok/s, planned July) both landed. Treat the LongCat training claim as vendor-asserted, but the direction is consistent: buyers should start asking where a model was trained and where it can be served as part of sovereignty and capacity planning.

Source: [SiliconANGLE; OpenAI](https://siliconangle.com/2026/06/30/chinas-meituan-open-sources-massive-longcat-2-0-ai-model-saying-trained-domestic-chips/).

## Benchmark moves

### Artificial Analysis Intelligence Index (v4.1)

Sonnet 5 debuted at 53 (max effort), +6 over Sonnet 4.6 and tying GPT-5.5 (high); Fable 5's July 1 reinstatement puts the index leader back on sale

  - Claude Fable 5 (max): 60 (64.9 at June launch under earlier eval set)
  - Claude Opus 4.8 (max): 56
  - GPT-5.5 (xhigh): 55
  - Claude Opus 4.7 (max): 54
  - Claude Sonnet 5 (max): 53 — new this week

Source: [Artificial Analysis](https://artificialanalysis.ai/models?intelligence=artificial-analysis-intelligence-index).

### SWE-bench Pro (vendor-scaffold aggregate)

Sonnet 5 entered at 63.2% (+5.1pt over Sonnet 4.6, with 85.2% on Verified); Fable 5's 80.0% score re-entered the available-model conversation over the prior ~69% ceiling

  - Claude Mythos 5: 80.3% — restricted (Glasswing only)
  - Claude Fable 5: 80.0% — available again 2026-07-01
  - Claude Opus 4.8: 69.2% — prior available leader
  - Claude Sonnet 5: 63.2% — new this week
  - GLM-5.2: 62.1% — open-weights leader

Source: [Anthropic system card via Vellum; benchlm.ai (updated 2026-07-03); vendor-scaffold numbers, not Scale-standardized](https://www.vellum.ai/blog/claude-sonnet-5-benchmarks-explained).

### LMArena

claude-sonnet-5-thinking added to the Code, Text, Search, Vision, and Document leaderboards on July 2; claude-fable-5 holds #1 overall while Elo settles

  - claude-fable-5: #1 overall (1653 in one view; 1509 in another scoring window)
  - GLM-5.2 (max): ~1584 — top open entry
  - claude-sonnet-5-thinking: ~1551 early snapshot; votes still accumulating

Source: [Arena leaderboard changelog](https://arena.ai/blog/leaderboard-changelog/).

### PutnamBench / miniF2F (formal math)

Leanstral 1.5 reset the open state of the art: miniF2F saturated at 100%, PutnamBench 587/672 — edging Seed-Prover 1.5 high by 7 problems at ~1/75th the per-problem cost

  - Leanstral 1.5 — miniF2F: 100% (val+test)
  - Leanstral 1.5 — PutnamBench: 587/672 at ~$4/problem
  - Leanstral 1.5 — FATE-H / FATE-X: 87% / 34% (SOTA)
  - Leanstral 1.5 — FLTEval pass@8: 43.2 vs Opus 4.6's 39.6

Source: [Mistral AI (vendor-reported; independent re-runs pending)](https://mistral.ai/news/leanstral-1-5/).

## Tier scorecard

_As of 2026-07-04._

| Tier | Leader | Challenger | Note |
|---|---|---|---|
| Closed frontier | Claude Fable 5 | Claude Opus 4.8 | Fable 5's July 1 restoration makes the board leader purchasable again — but it is metered to usage credits after July 7 and its 19-day outage proved gating risk, so keep Opus 4.8 as the tested fallback. |
| Open frontier | GLM-5.2 | LongCat-2.0 | GLM-5.2 keeps the permissive open lead on scores; LongCat-2.0 arrives with real OpenRouter demand evidence but staged weights — verify availability before piloting self-host. |
| Reasoning | Claude Fable 5 | GPT-5.5 | The AA Intelligence Index leader is available again; GPT-5.5 remains the strongest generally-available OpenAI entry while GPT-5.6 sits in gated preview. |
| Coding | Claude Fable 5 | Claude Sonnet 5 | Fable 5's 80.0% SWE-bench Pro is back on sale; Sonnet 5 (63.2% Pro, 85.2% Verified) is the cost-tier challenger — priced per task, not per token. |
| Multimodal | Gemini 3.5 Flash | MiniMax-M3 | No W27 multimodal reset; Google's media line churned (Veo 2.0/3.0 shut down June 30 as cheap successors shipped), which is a migration-budget signal rather than a leadership change. |
| Edge / small | Mellum2 | North Mini Code | No edge leadership change in-window; TwoTower's diffusion retrofit (2.42x throughput on a 30B-class backbone) is the efficiency pattern to watch for this tier. |

## Vendor signals

- **2026-06-26 · OpenAI + Anthropic (US Commerce) — Within five days the US government gated a new frontier launch (GPT-5.6, ~20 approved partners) and un-gated a suspended one (Fable 5 restored July 1)**
  - Frontier model availability is now partially a regulatory variable. Contracts and architectures should assume any flagship can be gated at launch or suspended post-launch; multi-model routing is availability insurance, not just cost optimization.
  - _Source:_ [OpenAI; Anthropic](https://openai.com/index/previewing-gpt-5-6-sol/)
- **2026-06-30 · Anthropic — Sonnet 5 launched at introductory $2/$10 per M tokens through August 31 (then $3/$15); Fable 5 moves to usage credits even on paid plans after July 7**
  - Budget owners should model post-promo Sonnet 5 economics now — the independent cost-per-task data already uses standard pricing — and treat Fable 5 as a premium metered resource, not a plan entitlement.
  - _Source:_ [Anthropic](https://www.anthropic.com/news/claude-sonnet-5)
- **2026-07-24 · DeepSeek — Legacy deepseek-chat / deepseek-reasoner aliases go fully dark July 24 15:59 UTC; they currently silently route to deepseek-v4-flash**
  - Anyone with DeepSeek integrations has a three-week migration deadline, and reasoner traffic wanting Pro-tier quality must explicitly move to deepseek-v4-pro. Open-weight vendors deprecate as aggressively as closed ones — budget for it.
  - _Source:_ [DeepSeek API docs](https://api-docs.deepseek.com/news/news260424)
- **2026-06-30 · Google — Veo 2.0/3.0 video models shut down June 30; same day, Nano Banana 2 Lite hit GA and Gemini Omni Flash entered public preview at $0.10/second of 720p video**
  - Google's media-model line is churning fast — aggressive price points on the way in, one-year-class deprecations on the way out. Media pipelines built on Gemini need version-migration budgets as a standing line item.
  - _Source:_ [Google Gemini API changelog](https://ai.google.dev/gemini-api/docs/changelog)
- **2026-07-01 · xAI — Voice Agent Builder beta: no-code speech-to-speech agents on Grok Voice at $0.05/min audio + $0.01/min telephony — roughly 1/3 to 1/5 of ElevenLabs/Vapi-class pricing**
  - Model vendors keep moving up-stack into the agent-platform layer, putting voice-AI middleware margins under direct attack from a model owner. Re-price any voice-agent build-vs-buy decision, and treat vendor-run voice benchmark claims as marketing.
  - _Source:_ [xAI](https://x.ai/news/grok-voice-agent-builder)
- **2026-06-30 · Anthropic — Claude Science beta shipped — a multi-agent scientific workbench on existing Claude models, with up to 50 funded projects at $30K credits each (applications through July 15)**
  - Anthropic is monetizing workflow ownership rather than raw capability, consistent with the W26 read that model adoption is turning into workflow operations. Expect vertical workbenches, not just model upgrades, to be the competitive surface in H2 2026.
  - _Source:_ [Anthropic; TechCrunch](https://www.anthropic.com/news/claude-science-ai-workbench)

## Watchlist

- **July 2026 — GPT-5.6 general availability.** Terra's claimed GPT-5.5-level capability at half the price is the biggest potential closed-model price reset of Q3; also watch whether the government preview process becomes a repeatable framework for every frontier launch.
- **July 2026 — Gemini 3.5 Pro GA.** A 2M-context flagship GA (slipped from June) would contest the closed-frontier ordering for the first time since May; it is already on LMArena and Antigravity for some users.
- **July 7, 2026 — Fable 5 usage-credit transition.** The included-allowance window ends; watch adoption and routing behavior once Fable 5 is purely metered, plus completeness of Bedrock/Vertex/Foundry re-enablement.
- **July 24, 2026 — DeepSeek legacy alias shutdown.** deepseek-chat / deepseek-reasoner go dark at 15:59 UTC; expect a burst of migration issues and possible V4 usage-share data.
- **Days to weeks — LongCat-2.0 full weight availability and independent benchmarks.** The HF repo said weights were coming soon at announcement (INT8/FP8 uploads observed since); independent evals will test both the near-frontier claim and the domestic-ASIC training story.
- **By August 31, 2026 — Sonnet 5 promo pricing expiry.** Standard $3/$15 pricing plus its measured token appetite could make Sonnet 5 more expensive per task than Opus 4.8 in production; re-run cost-per-task math before the promo lapses.
- **July 2026 — GPT-5.6 Sol on Cerebras.** Up to 750 tok/s serving for select customers would be the first datapoint for frontier closed-model serving on non-NVIDIA silicon at scale.

## Changelog

- Five tree rows added (claude-sonnet-5, gpt-5-6-sol, longcat-2-0, leanstral-1-5, nemotron-labs-twotower) and claude-fable-5 updated to record the July 1 restoration.
- Market architecture refreshed: five matching model cards, a July 2026 model-portfolio position read, and new W27 sources added.
- Excluded as unverifiable: the Pulpie family (Feyn Inc.) — aggregator mentions only, no primary source (grade 1-2).

---

Source of truth: `src/data/industry/models/2026-W27.ts`. Canonical HTML: <https://brianletort.ai/industry/models/2026-W27>. PDF: <https://brianletort.ai/downloads/model-pulse-2026-W27.pdf>. Tree: <https://brianletort.ai/industry/tree>.