RADAR ARCHIVE

Week 40, 2026

12 tools added to the directory and 6 launches tracked, Sep 28 – Oct 4, 2026.

Added to the directory (12)

Reviewed and written up.

OpenAI's week-later Sol update: near-Astra claims at one-fifth the price
Added Oct 1, 2026
GPT-6.1 Sol is OpenAI's September 29, 2026 release at DevDay, one week after GPT-6 Sol, positioned as the price-performance point of the GPT-6 line. OpenAI says it nearly matches GPT-6 Astra on agentic coding, computer use and professional work at one-fifth the standard token prices: $2 per million input and $10 per million output, with cached input cut to $0.10. The company reports the share of responses containing a factual error at low reasoning effort drops from 11.4% to 7.7%, staying within 1.9% of Astra across all reasoning settings, and that the model fails less often at flagging broken search tools, following explicit restrictions and avoiding unauthorized outcomes. It takes text and image input and returns text. It is available to Plus, Pro, Business, Enterprise and Edu users in...
Why: A week after GPT-6 Sol reset the coding price floor, GPT-6.1 Sol claims near-Astra agentic performance at the same $2/$10 rate — if those numbers hold under independent testing, this becomes the default for cost-sensitive agentic work. It enters the leaderboard on OpenAI's own claims, clearly labeled as such, with the scrapped GPT-6.1 Astra noted for context.
Paid Best Price-Performance Agentic Model Visit
Anthropic's mid-tier model that lands within a few points of Opus 5.5 at half the price
Added Sep 28, 2026
Claude Sonnet 5.5 is Anthropic's September 28, 2026 update to Sonnet 5 and the second model in the Claude 5.5 family after Opus 5.5. It keeps Sonnet 5's price of $2 per million input tokens and $10 per million output ($2.50 cache writes; cache reads were halved to $0.10 on October 7, 2026), generates output more than 30% faster, and in Anthropic's testing costs up to 30% less per task because it uses fewer tokens and tool calls. Anthropic's own benchmarks put it ahead of Opus 5.5 on Terminal-Bench 4.0 (70.6% vs 66.4%) and within two points of it on OSWorld 2.1 (80.1% vs 81.8%) and GDPval-AA v2.1 (1844 vs 1846). It is the first Sonnet to carry Opus-level cybersecurity safeguards, which route higher-risk security requests back to Sonnet 5.
Why: It is the rare mid-tier release that closes almost the whole gap to the flagship on agentic work while costing half as much per token, and Anthropic's customer reports (Box, Zendesk, Lovable, Base44) all point the same way: fewer steps and fewer tokens for the same result. For most Claude users it is the new default to try before paying for Opus.
Paid Best Value Everyday Claude Model Visit
Kling's next video model: 30-second clips, up to 15 references and 10 keyframes
Added Sep 29, 2026
Kling 4.0 is Kuaishou's next-generation Kling video model, announced September 28, 2026. It doubles the longest single generation to 30 seconds, continues a video across multiple shots up to 2 minutes, takes up to 15 references in text, image and video, and lets you pin up to 10 keyframes to control how a scene unfolds. Output goes up to 4K, with 10-bit HDR and a 21:9 ultra-wide ratio, plus stereo audio and more accurate lip sync. A lighter Kling 4.0 Flash went live the same day for Ultra Yearly subscribers only; Kling says the full Kling 4.0 is coming in October 2026.
Why: Clip length and control are what limit AI video for real storytelling, and Kling 4.0 moves both: 30 seconds per generation and keyframes to steer the action, rather than regenerating until something usable appears. Kling 3.0 is already one of the directory's top-ranked video tools, so this is the upgrade to watch. It is listed now with its limited availability stated plainly.
Paid Best Upcoming Long-Form Video Model Visit
Google's first Gemini 4 model: claimed top coding benchmarks, vetting-gated access
Added Oct 1, 2026
Gemini 4 Argon is the first model of Google's Gemini 4 generation, revealed September 30, 2026 by chief AI architect and Google DeepMind SVP Koray Kavukcuoglu — notably skipping a Gemini 3.5 Pro release. Per Google, it delivers frontier performance across real-world software engineering, enterprise knowledge work like legal and finance, and cybersecurity defense, scoring 77.9% on DeepSWE v1.1 against GPT-6 Astra, Claude Fable 5.1 and Opus 5.5, tying GPT-6 Astra on CWE-bench and leading the Vals Index. Its output ceiling rises to 1 million tokens, up from 64K. Access is limited to Google's Fairwind Program, serving vetted governments and cyber-authority customers, after a voluntary US government pre-release review; Google has given no general-availability date or pricing.
Why: Argon is the marker for where Google says the frontier now sits: claimed top scores on the coding benchmarks this directory's readers care about, plus a 16x larger output ceiling for long-horizon agent work. It ranks high on the strength of Google's own numbers, with access restrictions stated plainly — only vetted Fairwind organizations can use it today.
Enterprise Best Claimed Coding Benchmarks Visit
Meta's agentic coding model, tuned to finish long jobs in fewer steps
Added Sep 29, 2026
Muse Spark 1.3 is Meta Superintelligence Labs' September 2, 2026 update to its closed-weights Muse Spark model, released through the Muse Code harness (macOS and Linux) and the Meta Model API. Meta says it sustains longer multi-step work in a single thread, asks clarifying questions when stuck, and uses about 20% fewer tool calls and 25% fewer tokens than Muse Spark 1.2. API pricing is unchanged at $1.25 per million input tokens and $4.25 output ($0.15 cached input), with a cheaper Contributor tier at $0.10/$0.20 that lets Meta train on your data. A max reasoning setting, which produced Meta's best published scores, is now available in Muse Code and the API.
Why: On Artificial Analysis's Intelligence Index it scored 61, level with GPT-5.6 Sol and Claude Opus 5, according to VentureBeat, at a far lower per-token price than either. It is also the model behind Meta's Muse personal agent, so it is worth knowing even if you never call the API.
Paid Best Low-Cost Frontier Coding Model Visit
Fei-Fei Li's world model that turns a prompt or photo into a 3D space you can walk through
Added Sep 29, 2026
Marble is World Labs' generative world model. It builds persistent, navigable 3D environments from text, images, video, 360 panoramas or a rough 3D layout, then lets you edit parts of the scene, expand it, and combine several worlds into a larger one. Results export as Gaussian splats, collider meshes, high-resolution textured meshes on higher tiers, panoramas and rendered video. There is a free plan (up to 4 generations) and paid Standard, Pro and Max plans (up to 12, 25 and 75 generations), with commercial rights on Pro and Max. A separately billed World API launched in January 2026. AMD agreed on September 28, 2026 to acquire World Labs for about $8.2 billion in stock, with closing expected by the end of 2026.
Why: Most text-to-3D tools make a single object. Marble makes a whole space with consistent geometry you can move through, which is what game prototyping, virtual production and robot-training simulation actually need. It has a real free tier and exports to standard 3D formats rather than locking scenes into its own viewer.
Freemium Best for Whole 3D Environments Visit
Microsoft's preview image model, third on the editing board
Added Oct 2, 2026
MAI-Image-2.6 is Microsoft AI's image model for text-to-image and image editing. The model card dates the 2.6 release to 14 August 2026 and the faster MAI-Image-2.6-Flash release to 4 September 2026. It is a diffusion model with a maximum of 2,359,296 pixels, equivalent to 1,536 by 1,536, and either side can run past 1,536 if the total stays inside that cap. The 4 September post adds multi-image reference editing, web grounding, and dynamic aspect ratios, and puts both models in public preview on Microsoft Foundry and in the MAI Playground. Weights are not released.
Why: On 2 October 2026 it is third on Artificial Analysis AA-Image-Editing v2.0, at 1,137 Elo, and fifth on AA-Image-T2I v2.0, at 1,149 Elo. Microsoft's 4 September post said it was first for editing and second for text-to-image as of that day. The board has moved. It is still the strongest non-OpenAI model on the editing board, at a representative price of $38.90 per 1,000 images.
Paid Best for Editing Visit
Quality Mode in Grok for layout, type, and edits with up to five references
Added Oct 2, 2026
Grok Imagine Image 2.0 is xAI's image model, generally available 7 August 2026 as Quality Mode on grok.com/imagine and in the Grok iOS and Android apps. The launch post describes instruction following for dense layouts and small text, a magic wand and segmentation for local edits, background removal, smart resize across aspect ratios, and multi-reference generation with up to five input images. The API id is grok-imagine-image-2.0, on POST /v1/images/generations and POST /v1/images/edits.
Why: It is the highest text-to-image model on Artificial Analysis outside OpenAI. On 2 October 2026, AA-Image-T2I v2.0 has it at 1,157 Elo, rank 4, inside a rank band of 4 to 5. Editing is tenth, at 1,107 Elo, inside a band of 7 to 13. The model page lists output at $0.04 per image. The leaderboard prices the setting it scored at $60 per 1,000 images.
Paid Best for Typography Visit
Ideogram's edit model: change one thing, copy the rest of the pixels exactly
Added Oct 2, 2026
Ideogram 4.5 is Ideogram's image model announced 30 September 2026, built for edits that do not drift. On the Precise Edit endpoint, pixels the instruction does not touch are copied from the source, and the result comes back at that image's own width and height. Ideogram's model page shows this on a 4,016 by 6,016 pixel source. The same model also generates from text. The API takes a prompt, an optional mask, and up to four reference images (three when a mask is set), with quality set to very_low, low, medium, or high. It is live on ideogram.ai and at POST /v2/image/generate/ideogram-4-5 and POST /v2/image/precise-edit/ideogram-4-5. Open weights were promised and have not shipped.
Why: This is the Ideogram release that is actually about editing. Precise Edit is a specific guarantee, copied pixels and original dimensions, not a slogan, and an independent board already has a score for the High tier. It does not lead that board. It is the current Ideogram to reach for when the job is a series of small, local changes.
Freemium Best for Editing Visit
Open-source sandbox runtime that fences in what AI agents can touch
Added Sep 29, 2026
OpenShell is NVIDIA's open-source runtime for running AI agents inside a sandbox. It sits beneath the agent harness, runs the agent without privileges, limits which files it can reach, filters system calls at the kernel level, and checks every network request against policy by binary, destination, method and path. Nothing is allowed by default, credentials are only supplied where policy permits, a policy-aware router controls which model endpoints the agent can call, and every decision is logged. It supports Claude Code, Codex, GitHub Copilot CLI, Hermes, LangChain Deep Agents, OpenClaw and OpenCode, and runs on Docker, Podman, Kubernetes via Helm, and NVIDIA DGX Spark and DGX Station. First announced in March 2026, it became the software half of NVIDIA's Open Agent Safety Platform on...
Why: After a run of agents escaping their test environments, the useful question is where the fence lives. OpenShell enforces policy outside the agent's own process, where a prompt cannot talk it out of the rules, and it works with the coding agents people already use. It is free, open source under Apache 2.0, and needs no special hardware.
Free Best Open-Source Agent Sandbox Visit
A personal AI agent you text, which calls, books and buys on your behalf
Added Sep 29, 2026
Instinct is a personal AI agent that launched as an invite-only service in August 2026. You text it, and it uses its own phone number and computer to do the task: booking travel and restaurants, ordering groceries, paying bills, cancelling subscriptions and doing research. A concierge feature places phone calls to businesses that have no online booking, and a trusted person network lets your agent coordinate plans directly with friends' Instinct agents. It raised a $1 billion Series C from Sequoia, Benchmark and Coatue at a $10 billion valuation on September 28, 2026.
Why: It is the clearest independent rival to Meta's Muse among consumer agents that act rather than just answer, and its phone-calling concierge covers the offline businesses most agents cannot reach. It is listed as one to watch: access is still by invite, and the company has published no user numbers or pricing.
Free Best Agent for Phone and Text Errands Visit
Ideogram's first downloadable image model, built for layout and type
Added Oct 2, 2026
Ideogram 4.0 is Ideogram's first open-weight text-to-image model, released 3 June 2026. It is a 9.3 billion parameter flow-matching model trained from scratch, and it generates at any multiple of 16 from 256 to 2,048 pixels on a side, including native 2K. Prompts are structured JSON, with bounding boxes for layout and hex colors for the palette. A hosted magic-prompt step turns plain text into that JSON. The nf4 and fp8 weights are on Hugging Face under the Ideogram 4 Non-Commercial license, with inference code on GitHub. The same model is on ideogram.ai and the Ideogram API.
Why: This is the Ideogram generation you can download. On Artificial Analysis AA-Image-T2I v2.0, read 2 October 2026, the Quality tier is 1,011 Elo at rank 38 and the default tier is 1,005 Elo at rank 40. That sits level with Ideogram 4.5 on text-to-image, and well behind 4.5 on editing. Use 4.0 for design, in-image text, or a weight you can run. Use 4.5 when the edit has to leave the other pixels alone.
Freemium Best for Open-Weight Design Visit

On the radar (6)

Reported in the daily briefing this week. Not reviewed for the directory.

TechCrunch AI 2026-09-29

Meta Launches Muse for Small Business

Meta introduced a version of its Muse AI agent built for small businesses, connecting it to software such as Asana, Zoom, Intuit, Box, Canva, and Slack as well as a business's Meta ad accounts and Instagram and Facebook profiles. The tool is designed to help e…

Bloomberg Technology 2026-09-28

Nvidia Debuts System Meant to Stop AI Agents From Going Awry

Nvidia CEO Jensen Huang introduced a toolkit of software and hardware products that add independent security layers around AI agents, positioning it as an answer to a recent spate of rogue agent incidents. The company said its double layered security system wo…

TechCrunch AI 2026-09-28

Anthropic Releases Sonnet 5.5

Anthropic released Sonnet 5.5, which it says runs about 30 percent faster and consumes significantly fewer tokens than Sonnet 5, released three months earlier. Despite sitting below the flagship Opus line, Sonnet 5.5 outperforms Opus 5.5 on agentic coding task…