An entry point for models, market statistics, articles, papers, authors and standards. The editorial radar points to current primary sources; Frozen catalogs preserve snapshots, hashes, licenses, and inference limits.
Prices and availability below are declarations from suppliers themselves, reread in the primary source on August 13, 2026. Model price changes without warning — this radar is dated precisely because the date is part of the fact. FORGE treats model + harness + environment + eval as the decision unit; no supplier table proves universal superiority.
Highest capacity tier of the GPT-5.6 family. The supplier declares max and ultra such as effort levels; ultra coordinates four agents in parallel by default. Price unchanged from the 7/30 cut.
Balanced tier. OpenAI reduced price by 20% on July 30, 2026 — the previous cut of this radar published the launch values.
Lowest cost tier. OpenAI reduced price by 80% on July 30, 2026; the value published in the previous cut was five times higher than that declared today.
Mythos-class model for long projects. Anthropic states additional safeguards, automatic routing to less capable models in sensitive domains, and 30-day retention for security monitoring.
Claude family general purpose tier 5. Supplier offers fast mode in research preview for Opus 5, charged at US$10 / US$50 per 1M.
Balanced tier. The introductory price became the standard: The increase to $3 / $15 scheduled for September 1, 2026 has been canceled by the supplier.
Google describes it as its smartest model built for speed. There are Batch and Flex tiers at half price and Priority at 1.8×.
Stated price per prompt range: Over 200k input tokens, xAI charges $4 / $12. Cached input costs $0.50.
Market statistics change quickly. Each number below carries source, population and year; it informs FORGE's design, but does not replace the product's own metrics.
The report also records 362 documented AI incidents, up from 233 in 2024, and 53% population adoption of generative AI in three years.
Among 13 agents with border autonomy, only four disclosed any agentic security assessment; 20 out of 30 supported MCP.
In Claude.ai's November 2025 clip, augmentation once again predominated. In API traffic, computing and math tasks reached 46%.
DORA 2025 positions AI as an amplifier of organizational strengths and weaknesses. The greatest return comes from the work system, not from the isolated tool — a foundation compatible with FORGE's thesis.
Individual authorship appears when the official source declares it. Specs, standards, and collective reports remain attributed to the responsible institution — without inventing an author.
Report of an agent-first product: knowledge of the repository as a system of record, isolated worktrees, readable feedback and mechanically applied architectural invariants.
Describes initializer, incremental progress, clean state, and end-to-end verification. The article credits contributions from David Hershey, Prithvi Rajasakeran and other members of the code RL and Claude Code teams.
Specifies orchestration, isolated workspaces, work tracking, and observability. FORGE uses the source as design input, not as proof of compliance.
AI Index Steering Committee, seven AI Agent Index experts, and the DORA program form an institutional foundation for performance, transparency, adoption, and sociotechnical systems.
System cards describe supplier evaluations. NIST, OWASP, OpenSSF, and OpenTelemetry add standards and controls outside the model. No layer replaces the others.
The list below mirrors the 40 entries from the two public catalogs. For each cut, the source catalog records publisher, class, independence group and URL; manifests and receipts add policy, snapshot and hash.
The July 17 cutoff supports the v1.7 benchmark; the one from August 4th supports FORGE 3.1. The model editorial radar above is more recent and remains separate from these receipts.
How to read this: 40 receipts prove acquisition and source binding in the aforementioned cuts. They do not automatically attest to full source veracity, fleet adoption, customer results or operational maturity.
Trustyu Forge