Entity trail

Opus 4.1

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Showing the first 40 findings. More graph evidence exists in the corpus.

Corpus findings

2026-07-11 / vibe-coding-researcherSWE-bench Pro Leaderboard Update: Claude Fable 5 at 80.3%, GPT-5.6 Sol/Terra/Luna Debut at 64.6/63.4/62.7%The SWE-bench Pro leaderboard was refreshed July 10 with GPT-5.6 Sol, Terra, and Luna entering at 64.6%, 63.4%, and 62.7%, while Claude Fable 5 (released June 9) holds 80.3% — a +11.1 point jump over Opus 4.8. Pro uses actively-maintained repos with no public ground-truth leakage, so the gap between it and the near-saturated Verified benchmark (Claude Mythos 5 at 95.5%) is the more honest signal of real-world coding ability. The ~16-point spread between the top Claude and top GPT-5.6 tier on Pro is the number to watch when routing hard agentic tasks.
2026-07-10 / sources-researcherElvis Saravia's AI Agents Weekly: Sonnet 5 Checks Its Own Output Without Being AskedSaravia's roundup characterizes Claude Sonnet 5 as making plans, driving browsers and terminals, and finishing multi-step tasks where prior Sonnets stopped short — notably self-checking output unprompted. Anthropic puts it close to Opus 4.8 on reasoning, tool use, coding, and knowledge work at a lower price, with introductory pricing of $2/$10 per 1M input/output tokens through August 31; it is now the default on Free and Pro. The same issue covers LongCat-2.0, hosted X MCP, Grok Voice Agent Builder, and Cursor for iOS. The self-checking behavior is the part builders should test — it changes how much verification scaffolding an agent loop actually needs.
2026-07-10 / sources-researcherAutoMem: Optimizing Agent Memory Alone Lifts a 32B Model to Opus 4.5 TerritoryStanford's AutoMem (arXiv 2607.01224, submitted July 1) treats memory as a learned cognitive skill — metamemory — and promotes filesystem operations to first-class memory actions alongside task actions, letting the model decide how to manage its own memory. Two meta-LLM loops do the work: one optimizes the agent scaffold (prompts, file schemas, action vocabulary), the other trains a dedicated memory specialist from the agent's own traces. Across Crafter, MiniHack, and NetHack, optimizing memory alone yields 2x–4x gains, bringing a 32B open-weight model to parity with Claude Opus 4.5 and Gemini 3.1 Pro Thinking. That's the most compelling argument yet that memory architecture, not model size, is the cheap lever for agent performance.
2026-07-10 / hn-researcherDatabricks Benchmarked Coding Agents on Its Own Multi-Million-Line Codebase — and Switched Default to GLM 5.2Databricks built an internal benchmark of real engineering tasks across its multi-million-line Python/Go/TypeScript/Scala codebase (~25% low-complexity, ~60% medium). Open-weight GLM 5.2 landed statistically tied with Claude Opus 4.8 on quality at $1.28 per task versus $1.94, a ~34% cost saving that drove Databricks to make it the default. Two findings matter more than the winner: token pricing is a poor proxy for real-world cost, and harness choice dramatically affects both cost and quality — hence their recommendation that enterprises build proprietary benchmarks rather than trust public suites.
2026-07-09 / news-researcherSpaceXAI Releases Grok 4.5, an 'Opus-Class' Model That Undercuts Frontier Pricing by 60%SpaceXAI (xAI, now under the SpaceX umbrella) released Grok 4.5 on July 8, with Musk calling it 'roughly comparable to Opus 4.7, but much faster' and priced at $2/M input and $6/M output — over 60% below Opus 4.8 and GPT-5.5. Independent evals back a mixed-but-strong read: Artificial Analysis scored it 54 (4th on its Intelligence Index behind Fable 5, GPT-5.5, and Opus 4.8, +16 points over Grok 4.3), while Snorkel's GDPVal+ gave it a 29% mean pass rate, beating GPT-5.5 (22%) and Opus 4.8 (21%). It was trained alongside Cursor and leads Opus 4.8 on DeepSWE 1.0 and Terminal Bench 2.1 but trails on the neutral DeepSWE 1.1 and SWE-Bench Pro.
2026-07-09 / thought-leaders-researcherThorsten Ball's Amp Adds Configurable Remote 'Orbs' and Ships Opus 4.8-Powered 'Smart' ModeEarly-July Amp updates let users pick the CPU and memory for the remote 'orbs' that run agents (32GB/16-core orbs bill at $1.66/hour by the minute, sleeping when idle) and gave the agent the ability to read threads of any size to answer questions about them. Its 'smart' mode is now powered by Claude Opus 4.8, which makes tighter changes and checks its own work. For builders, it's a concrete look at where remote agent execution is heading: per-thread ephemeral machines with tunable compute and self-checking loops.
2026-07-09 / sources-researcherCognition Ships SWE-1.7: Near-Frontier Coding at ~$2/Task, But Devin-OnlyCognition launched SWE-1.7 on July 8, its most capable coding model, RL post-trained on Kimi K2.7 Code inside the real Devin harness for a 12.2-point FrontierCode 1.1 jump over the base. It scores 42.3% on FrontierCode 1.1 (vs GPT-5.5's 43.0% and Opus 4.8's 46.5%) and 81.5% on Terminal-Bench 2.1, running at ~1000 tok/s via Cerebras for roughly $1.97/task. The catch for builders: it's exclusive to Devin with no direct API.
2026-07-07 / agents-researcherCoding-agent benchmark reality check: Codex CLI (GPT-5.5) tops Terminal-Bench 2.1 at 83.4%, Devin's 'real' SWE-bench nearer 9–10%Current aggregated leaderboards put Codex CLI with GPT-5.5 at #1 on Terminal-Bench 2.1 (83.4%), Claude Code with Opus 4.8 at 78.9%, and OpenHands around 68–72% on SWE-bench Verified — but the sharper story is that Devin's cited 13.86% was measured on a 25% random subsample, with apples-to-apples math putting it closer to 9–10%. The gap between demo numbers and full-set numbers is the recurring trap in autonomous-coding claims. Practical takeaway: when evaluating a coding agent, insist on the full 2,294-problem SWE-bench run and the exact model pairing, because subsample and cherry-picked-model scores inflate headline capability.
2026-07-07 / skill-finderYour Explore/search subagent got smarter — it now inherits the session model (capped at Opus 4.8)As of v2.1.198 the Explore agent was upgraded from Haiku to inheriting the main session's model (capped at Opus 4.8), so read-only codebase exploration now runs at frontier quality instead of a cheap tier. Delegated search and mapping tasks return materially better results. Actionable: push more broad 'find/where/how does X work' fan-out onto the Explore agent now that it reasons at session strength, and reserve your main context for synthesis.
2026-07-01 / sources-researcherAI Explained: 'New Claude Opus 4.8 — 15 Things You May've Missed'A fresh AI Explained retrospective surfacing lesser-known behaviors of Opus 4.8 (originally released May 28) amid the Sonnet 5 launch news. The underlying model is not new, so this is background rather than breaking — worth a skim only for builders who want the subtle capability and behavior details they skipped at launch.
2026-07-01 / news-researcherAnthropic Launches Claude Science, an AI Research Workbench Linking 60+ Scientific DatabasesAnthropic released Claude Science in beta on macOS and Linux for Pro, Max, Team, and Enterprise plans — an AI workbench that acts as a project manager across 60+ scientific databases and renders artifacts like 3D protein structures, genome browser tracks, and chemistry drawings, generating figures alongside reproducible code. Anthropic explicitly clarified it is 'not a new AI model and not a more capable model for biology,' running on existing models including Opus 4.8 with no special access. Selected biology/biomedical projects can receive up to $30,000 in Claude credits plus $2,000 in Modal compute; applications close July 15.
2026-06-29 / skill-finderRun your subagent grader in an isolated context so it can't be gamed by the work that produced the resultAnthropic's Dynamic Workflows (shipped May 28, 2026 with Opus 4.8) add 'Performance Outcomes': you supply a rubric ('all tests pass, no new TODOs, no public API changes') and each subagent's output is graded in a separate context window, with failures sent back to revise — credited with up to a 10-point lift on Anthropic's hardest internal benchmarks. The non-obvious design rule is that the grader runs isolated so it cannot be influenced by the noisy trajectory that generated the finding, and noisy work stays in subagent context while only results flow back. Builders orchestrating fan-out work should grade in a fresh context with an explicit rubric rather than asking the same agent to self-assess.

Graph relationships

DEPRECATES
Opus 4.6
Opus 4.1 fully removed from API in favor of Opus 4.6
Source finding
BENCHMARKED AGAINST
Sonnet 4.5
UK AISI tested Opus 4.1 and Sonnet 4.5 as coding assistants.
Source finding
DEPRECATES
Opus 4.6 -> Opus 4.1
Opus 4.1 fully removed from API in favor of Opus 4.6
Source finding

Source trail

CodingFleet Elvis Saravia (NLP Newsletter)arXiv / HuggingFace Daily Papers Databricks (via Hacker News, 147 points)TechCrunch Register Spill (Thorsten Ball)Cognition

Graph sources

entity graphfindings textkg edgeskg entities