// live intelligence desk — updated 24 Jul 2026

Every frontier model,
tracked on one screen.

Signal is Securet's independent desk on the AI industry — release timelines, lab-by-lab moves, and a live news feed, read in minutes instead of chased across twelve blogs.

9labs tracked
24models this month
1Mtokens — new context ceiling
Live

Latest AI & security headlines

Fetched client-side from public RSS feeds — cached in-session, auto-refreshes every 8 minutes while this tab is open.

Connecting to feeds…

Sources: TechCrunch AI, VentureBeat AI, MIT Technology Review AI, Hacker News. Headlines link to the original publisher — Signal doesn't rehost article content.

Release log

What shipped, in order

Every frontier and near-frontier release since late June — newest first.

Date
Model
Lab
Status
Jul 21
Gemini 3.5 Flash LiteLightweight tier of the 3.5 Flash line
Google
new
Jul 16
Kimi K3Moonshot AI's latest open-weight release
Moonshot AI
new
Jul 9
Muse Spark 1.1Agentic + computer-use focus, 1M-token context
Meta
shipped
Jul 9
GPT-5.6 Sol / Terra / LunaThree-track flagship line, 1M-token context
OpenAI
shipped
Jul 8
Grok 4.5Default for SuperGrok/X Premium+, strong real-time web context
xAI
shipped
Jul 1
Claude Fable 5Returned after a 19-day export-control pause
Anthropic
shipped
Jun 30
Claude Sonnet 5Fast general-purpose tier, 1M-token context
Anthropic
shipped
Jun 16
GLM-5.2Open-weight release
Z.ai
shipped
Jun 9
Claude Fable 5 & Mythos 5Mythos limited to approved organizations only
Anthropic
shipped
May 28
Claude Opus 4.8Top-ranked overall on Intelligence Index at launch
Anthropic
shipped
Lab tracker

Who's building what

The nine labs shipping frontier or near-frontier models this cycle.

United States · Anthropic
Anthropic

Claude 5 generation — Sonnet 5 for everyday work, Opus 4.8 as the top-ranked overall model, Fable/Mythos 5 for the highest tier.

Latest: Sonnet 5, Jun 30 2026
United States · OpenAI
OpenAI

GPT-5.6 three-track flagship line (Sol, Terra, Luna) plus GPT-Live voice, now the new ChatGPT default.

Latest: GPT-5.6, Jul 9 2026
United States · Google
Google DeepMind

Gemini 3.1 Pro leads on hardest-mode reasoning; the 3.5 Flash line is the price-performance frontier pick.

Latest: Gemini 3.5 Flash Lite, Jul 21 2026
United States · Meta
Meta AI

Muse Spark line pushes agentic work and computer-use with a 1,000,000-token context window.

Latest: Muse Spark 1.1, Jul 9 2026
United States · xAI
xAI

Grok 4.5 is the default SuperGrok/X Premium+ model, strong on real-time web and X context.

Latest: Grok 4.5, Jul 8 2026
China · DeepSeek
DeepSeek

V4 line (Pro and Flash) continues to close the gap with closed frontier labs on reasoning benchmarks.

Latest: DeepSeek-V4-Pro, Apr 24 2026
China · Moonshot AI
Moonshot AI

Kimi K-series open-weight models are shipping on a fast cadence, K3 arriving mid-July.

Latest: Kimi K3, Jul 16 2026
France · Mistral
Mistral AI

Medium-tier models remain the go-to for teams wanting strong open-weight performance in the EU.

Latest: Mistral Medium 3.5, Apr 29 2026
China · Z.ai
Z.ai

GLM line continues iterating quickly on open weights aimed at agentic and coding workloads.

Latest: GLM-5.2, Jun 16 2026
The briefing

Three shifts worth knowing this month

01

The gateway layer became the security layer

Portkey open-sourced its full gateway under Apache 2.0, then was folded into Palo Alto Networks' Prisma AIRS in a May 2026 acquisition — routing, guardrails, and agentic-AI security are converging into one category.

02

1M-token context is now the baseline

GPT-5.6, Claude Sonnet 5, and Meta's Muse Spark 1.1 all shipped with million-token context windows within weeks of each other — a 2025 differentiator is now table stakes.

03

Agents moved from demo to daily tool

ChatGPT Work, Claude Cowork, Gemini's managed agents, and Microsoft's Sales/Service Agents all launched or expanded — the fight is now "whose agent finishes the task," not "best chat model."

Snapshot

Intelligence Index — top of the field

A composite reasoning/knowledge score reported across providers. Rankings shift with every release — treat as directional.

#
Model
Score
1
Claude Opus 4.8 — Anthropic
61
2
GPT-5.6 line — OpenAI
~60
3
GPT-5.5 — OpenAI
59-60
4
Gemini 3.1 Pro — Google
57
4
Qwen 3.7 Max — Alibaba
57
6
Gemini 3.5 Flash — Google
55
Two products, one desk

One gateway. Your own API keys. Every lab on this page.

Signal is the research desk. The Securet AI Control Plane Gateway is the product it feeds — bring your own OpenAI, Anthropic, Google, Meta, and xAI keys, and get unified routing, cost optimization, and guardrails from one endpoint.

Live from the gateway

Model routing & pricing

Pulled live from GET /v1/models on your deployed gateway — no separate pricing page to keep in sync by hand.

Connecting to gateway…
Signal — an AI intelligence desk by Securet, ai.securet.com.au
Data compiled Jul 24 2026 · figures are directional, re-verify before procurement decisions
// detection & response

AI SOC Console

Paste a raw alert, log line, or SIEM export. The engine extracts indicators, maps activity to MITRE ATT&CK tactics, scores severity, and generates a first-response playbook — entirely in your browser. Nothing leaves this page unless you use the optional AI deep-dive below.

📥 Input

Tip: Ctrl/Cmd + Enter runs detection from the text box.

🎯 Detection summary

Run detection to see severity, ATT&CK tactics, and extracted indicators here.

🤖 AI deep-dive (optional)

Sends the alert + detection summary through your Securet AI Control Plane Gateway for a written incident narrative and escalation recommendation. The gateway picks the model (routing rules below), holds your provider keys server-side, and this page never sees or sends them — only your gateway key, which is scoped and revocable.

⚠ This calls the gateway URL above from your browser — update it once your gateway is deployed (see the gateway README). Until then this will fail with a connection error, which is expected. Your gateway key is still visible in this browser's network requests (same as any API key typed client-side), but the underlying OpenAI/Anthropic/Google/etc. keys it maps to never leave your server.

AI SOC Console — a Securet tool. Local heuristics only; not a replacement for a licensed SIEM/SOAR platform.
// self-serve

Dashboard

Get a gateway key, add your own provider keys, and check usage — this page talks directly to your deployed gateway, no separate account system here.

🆕 New here? Get a gateway key

Free tier: $5/month budget, 20 requests/min — raise this later via the gateway's admin API.

🔑 Already have a key? Check usage

Enter your gateway key to see spend and recent calls.

➕ Add a provider key (BYOK)

Encrypted at rest on your gateway server. Never stored in this browser, never sent anywhere except your own gateway URL above.