Frontier AI coding capability reaches new heights as Claude Fable 5 reclaims public availability and OpenAI's GPT-5.6 family edges toward broad release, while hyperscaler capex commitments collectively cross $725bn for 2026.
Frontier Model Releases
Anthropic Claude Fable 5Mythos-class / Returned to global access
Claude Fable 5, Anthropic's first publicly available Mythos-class flagship, returned to global access on July 1, 2026 after an 18-day government-mandated suspension related to export-control jailbreak concerns. It leads the SWE-bench Pro leaderboard at 80.3% and is purpose-built for long-horizon agentic runs at $10/$50 per 1M tokens. Safety classifiers on cyber/bio/chemistry queries fall back to Opus 4.8. — 1 Jul · felloai.com, codingfleet.com, buildfastwithai.com, VentureBeat
Anthropic Claude Sonnet 5New free & Pro default
Released June 30, 2026, Claude Sonnet 5 became the new free and Pro default on claude.ai, leading Artificial Analysis's professional-writing benchmark. It ships with a 1M-token context at introductory pricing of $2/$10 per 1M tokens (rising to $3/$15 after August 31, 2026). It beats Claude Opus 4.8 on Terminal-Bench while offering roughly 40–60% lower cost. — 30 Jun · aireleasetracker.com, felloai.com, llmgateway.io
OpenAI GPT-5.6 (Sol / Terra / Luna)Gated preview — not yet GA
Previewed June 26, 2026 to approximately 20 government-approved partner organizations following a Trump executive order mandating a 30-day pre-release federal review. Sol is the flagship coding tier at $5/$30 per 1M tokens; Terra is balanced at $2.50/$15; Luna is the fast volume tier at $1/$6. General availability is expected mid-July 2026, with Sol deploying on Cerebras infrastructure at up to 750 tokens per second. — 26 Jun · VentureBeat, felloai.com, buildfastwithai.com, designforonline.com
xAI Grok 4.5Private beta only
Elon Musk confirmed on June 28, 2026 that Grok 4.5 is running in private beta with teams at SpaceX and Tesla, built on a fresh V9 foundation with roughly 1.5 trillion parameters. No public release date has been set; Grok 4.3 remains xAI's public flagship at $1.25/$2.50 per 1M tokens with a 1M-token context window. — 28 Jun · felloai.com, designforonline.com, buildfastwithai.com
Kimi Kimi K2.7 CodeOpen weights / Modified MIT
Released June 10–12, 2026 by Moonshot AI, Kimi K2.7 Code is a trillion-parameter MoE coding model with a 262K context window. It is positioned for agentic coding with native terminal and tool use, priced at $0.95/$4.00 per 1M tokens, and cited as the newest coding-first open-weight leader with a clear commercial license. — 12 Jun · autom8labs.io, buildfastwithai.com, llmgateway.io
Google Gemini 3.5 FlashProduction / Free default
Gemini 3.5 Flash is positioned as Google's fast, production-friendly default model in the Gemini app and Google AI Studio, serving as the free-tier default. It is cited as the price-performance runner-up in coding behind Claude Fable 5 and the primary budget alternative across frontier model comparisons in July 2026. — Jun · felloai.com, buildfastwithai.com, blog.mean.ceo
Meta Muse SparkMultimodal / Creative
Meta's Muse Spark has been released and is positioned for creative and multimodal tasks. CEO Mark Zuckerberg offered no firm schedule for any subsequent improved models following the launch. Meta Superintelligence Labs also confirmed its first closed-weight, natively multimodal reasoning model, marking a pivot away from open Llama releases. — Jun · Tom's Hardware, outyet.ai, felloai.com
xAI Grok 4 Open 100B-A20BOpen weights / Custom license
xAI released its first open-weights model, Grok 4 Open 100B-A20B, under a custom license (not Apache/MIT — commercial use terms require independent review). This marks xAI's first open-weight contribution to the ecosystem, though it carries more restrictive terms than MIT- or Apache-licensed alternatives. — Jun · autom8labs.io
Nvidia unveils Rubin platform; Blackwell Ultra delivers 50x agentic AI perf vs Hopper
NVIDIA's FY2026 Q4 filing reveals the Rubin platform comprising six new chips targeting up to 10x inference token cost reduction vs Blackwell. Blackwell Ultra delivers up to 50x better performance and 35x lower cost for agentic AI vs Hopper per SemiAnalysis InferenceX benchmarks. AWS, Google Cloud, Azure, and OCI are first-to-deploy. — NVIDIA 8-K (FY2026), SemiAnalysis
Nvidia announced an $80bn share buyback program. A contested report suggests the next-generation Kyber AI server rack launch has slipped to 2028, which Nvidia denied. Gross margin rose to 74.1% on continued hyperscaler GPU demand. — Yahoo Finance / NVDA Q filings
Cloud
Google Cloud Q1 2026 revenue +63% YoY to $20bn; backlog reaches $462bn
Google Cloud reported Q1 2026 revenue of $20.03bn, up 63% year-over-year, beating consensus. Cloud contracted backlog nearly doubled sequentially to $462bn. However, $190bn CY2026 capex guidance cut free cash flow margin from 21% to 9.2%, pressuring GOOGL shares. — Alphabet Q1 2026 earnings, TradingKey
Compute
Chinese AI models gaining enterprise traction in US as cost advantage widens
CNBC reported July 7, 2026 that recent releases from DeepSeek and Z.ai are being adopted by US enterprises as OpenAI and Anthropic costs surge. DeepSeek maintains a permanent 75% pricing discount on its API, with DeepSeek V4-Flash at $0.14/$0.28 per 1M tokens. — CNBC (7 Jul 2026), felloai.com
Infrastructure & Capital
Compute
OpenAI ↔ Nvidia — 10GW, up to $100bnup to $100bn
OpenAI and Nvidia announced a letter of intent for a landmark strategic partnership to deploy at least 10 gigawatts of Nvidia systems for OpenAI's next-generation AI infrastructure. Nvidia intends to invest up to $100bn in OpenAI progressively as each gigawatt is deployed. The first gigawatt on the Vera Rubin platform is targeted for H2 2026. — NVIDIA Newsroom (official press release)
Funding
OpenAI closes $122bn round at $852bn valuation — largest private financing in history$122bn
OpenAI closed $122bn in committed capital at an $852bn post-money valuation on March 31, 2026. The round was anchored by Amazon ($50bn, of which $35bn is contingent on IPO or AGI milestone), Nvidia ($30bn), and SoftBank ($30bn), co-led with a16z, D.E. Shaw, MGX, TPG, and T. Rowe Price. OpenAI now generates $2bn/month in revenue and serves 900 million weekly ChatGPT users. A confidential S-1 was filed with the SEC on June 8, 2026; IPO timing now reportedly targets 2027. — Bloomberg (31 Mar 2026), OpenAI blog (official), TechCrunch
Compute
Stargate JV — $500bn target through 2029; Abilene campus live at 1.2GW$500bn target
The OpenAI/SoftBank/Oracle Stargate joint venture targets $500bn in AI data center investment over four years, announced January 2026. By April 2026, the flagship Abilene, Texas campus was live with 1.2GW of Oracle Cloud Infrastructure. The program has expanded to nearly 7GW of planned capacity and over $400bn in committed investment across Michigan, Wisconsin, Wyoming, Pennsylvania, and multiple Texas locations. — TechCrunch (Feb 2026), IntuitionLabs / Oracle-OpenAI analysis
Compute
Hyperscalers — $725bn combined CY2026 capex, up 77% from 2025$725bn combined
Google, Amazon, Microsoft, and Meta collectively plan to spend $725bn on capex in 2026, up 77% from the prior year's record $410bn, per Q1 2026 earnings compiled by the Financial Times. Amazon leads at $200bn, followed by Microsoft and Alphabet each at ~$190bn, and Meta at $115–135bn. Analysts project combined capex could top $1tn in 2027. — Yahoo Finance / Financial Times Q1 2026 earnings compilation, Tom's Hardware
Funding
Together AI — $800M Series C led by Aramco Ventures$800M
Together AI closed $800M in Series C funding on July 1, 2026, led by Aramco Ventures with Nvidia, Vista Equity Partners, and General Catalyst participating. Annual bookings crossed $1.15bn in Q2 2026. The company plans to grow public cloud inference capacity by 50x over five years, running open-source models in a serverless environment. — StartupHub.ai (6 Jul 2026), Crunchbase
Compute
Crusoe in talks for $3bn raise — valuation would triple to ~$18bn$3bn (in talks)
GPU cloud provider Crusoe reported it was in talks to raise $3bn, a round that would triple its valuation to roughly $18bn. The raise follows Crusoe's role in the Stargate Abilene expansion and signals continued private capital demand for dedicated AI inference and training cloud infrastructure. — StartupHub.ai (6 Jul 2026)
Compute
Kling AI — $2bn close at $18bn valuation (General Atlantic)$2bn
Kling AI, the Chinese video AI company, closed a $2bn funding round backed by General Atlantic at an $18bn valuation. The close highlights continued international capital flows into specialized AI content generation platforms, sitting alongside the broader infrastructure buildout. — StartupHub.ai (6 Jul 2026)
Funding
Abu Dhabi MGX Fund I closes at $49bn; 3GW Europe AI campus under development$49bn fund
Abu Dhabi's MGX closed Fund I at $49bn, exceeding its $45bn target, with investments across 14 companies spanning semiconductors, AI infrastructure, and AI platforms. MGX is also developing Europe's largest AI campus near Paris with 3GW compute capacity and acquired Aligned Data Centres as part of a $40bn consortium deal. — crescendo.ai (Jul 2026)
Compute
Nvidia ↔ Meta — multiyear multigenerational GPU partnership including millions of Blackwell and Rubin GPUsundisclosed
Nvidia's FY2026 Q4 filing revealed a multiyear, multigenerational strategic partnership with Meta spanning on-premises, cloud, and AI infrastructure, including the large-scale deployment of Nvidia CPUs, networking, and millions of Blackwell and Rubin GPUs. Meta has earmarked $115–135bn for 2026 capex and plans $600bn in US infrastructure through end of 2028. — NVIDIA 8-K (FY2026), TechCrunch (Feb 2026)
Not verified today: SWE-bench Verified scores for GPT-5.5, GPT-5.6, Gemini 3.1 Pro, and Grok 4.3 could not be independently verified from a dated primary source as of 8 Jul 2026; the leaderboard above reflects only scores sourced from BenchLM.ai and digitalapplied.com · Exact Alphabet (GOOGL) market cap as of 7–8 Jul 2026 not confirmed from a single real-time primary source; ~$2.9tn cited from Motley Fool aggregation · Microsoft exact CY2026 capex figure is contested across sources ($145bn annualized run-rate per Yahoo Finance Feb 2026 vs $190bn per Q1 2026 FT compilation); $190bn used as latest figure · Mistral had no new model release verified in the last 7 days; excluded per brief instructions · DeepSeek R2 / V4-Thinking: long-rumored reasoning flagship — still unannounced as of 8 Jul 2026 per outyet.ai · Grok 4.5 benchmark scores not publicly available; model remains in private beta · GPT-5.6 (Sol/Terra/Luna) benchmark scores not independently verified; model not yet generally available · Meta Muse Spark detailed benchmark scores not confirmed from a primary source · Nvidia NVDA share price and market cap on 8 Jul 2026 specifically (market opens after research cutoff); closest available is 6 Jul 2026 at ~$4.72tn · OpenAI IPO timing and S-1 details remain unconfirmed; company stated listing 'may be a while' and is reportedly leaning toward 2027
Sources: Bloomberg (31 Mar 2026 — OpenAI $122bn round) · OpenAI blog — official funding announcement · TechCrunch — hyperscaler capex & Stargate (Feb/Mar 2026) · NVIDIA Newsroom — OpenAI 10GW partnership press release · NVIDIA 8-K FY2026 (SEC filing) — Rubin platform, Blackwell Ultra, Meta partnership · Yahoo Finance / Financial Times — $725bn hyperscaler capex compilation (Q1 2026 earnings) · VentureBeat — GPT-5.6 limited preview & US executive order · CNBC (7 Jul 2026) — Chinese AI model adoption · BenchLM.ai — SWE-bench Verified leaderboard (updated 6 Jul 2026) · morphllm.com — SWE-bench Pro leaderboard analysis (Jun 2026) · digitalapplied.com — SWE-bench scaffolding analysis (Jun 2026) · vals.ai — independent Fable 5 score confirmation · codingfleet.com — SWE-bench Pro leaderboard (updated Jun 9, 2026) · felloai.com — Best AI Models July 2026 · buildfastwithai.com — Best AI Models July 2026 ranked · aireleasetracker.com — AI model release tracker · llmgateway.io — LLM release timeline · autom8labs.io — AI Insight July 2026 · StartupHub.ai — AI infrastructure week Jul 6, 2026 · Crunchbase — weekly funding roundup Jun 27–Jul 2, 2026 · crescendo.ai — VC investment deals AI 2026 · StockAnalysis.com — NVDA/MSFT market cap (Jul 2026) · CompaniesMarketCap.com — NVDA market cap (6 Jul 2026) · Macrotrends — MSFT/NVDA market cap historical · TradingKey — Alphabet Q1 2026 earnings analysis · outyet.ai — AI model release tracker · designforonline.com — Best AI models 2026 roundup · tom's hardware — hyperscaler capex (Apr 2026)
