Claude Opus 4.6 First Look: A Million Tokens of Memory, Four Speeds of Thinking
The gate: Five stars need 21+ and Australian Fit of at least 4/5.
Skip to the verdictLaunched: 5 February 2026 · Reviewed: 6 February 2026 · Score: 20/25 ⭐⭐⭐⭐ Very Good
Verdict in one line: The strongest model Anthropic has shipped — and the first whose context window is big enough to hold an entire Australian small business's paperwork in one conversation.
📌 Newer models available — updated 20 September 2026. This first look covers Claude Opus 4.6 as it launched in February 2026. Anthropic's line-up is now Fable 5.1, Opus 5, Sonnet 5 and Haiku 4.5; Opus 4.6 is a legacy model. For our current, fully-scored assessment of Claude, read the full Claude review. We've kept this first look in place as a dated record of what Opus 4.6 introduced.
⚠️ STALENESS FLAG — 5 September 2026. This first look was published 6 February 2026 and is 211 days old. It has not been re-verified against the vendor since. AI products move fast: pricing, model versions, plan names and Australian availability may all have changed. Read the verdict below as a snapshot of February 2026, not as a current recommendation. (House rule: any review over six months old carries this flag until it is updated.)
What launched
On 5 February 2026 Anthropic released Claude Opus 4.6, its new flagship model. The headline is a 1-million-token context window (in beta) — roughly 700,000 words in a single conversation. Anthropic reports state-of-the-art results on Terminal-Bench 2.0, Humanity's Last Exam, and GDPval-AA, the benchmark that measures performance on real knowledge work.
What it does
Two changes matter more than the benchmarks. First, adaptive thinking with four effort levels — you choose low, medium, high, or max, trading speed against depth per task. A quick email summary doesn't need the same gear as a contract analysis, and now it doesn't cost the same time. Second, context compaction quietly squeezes long conversations so they keep going instead of hitting a wall. Output stretches to 128,000 tokens — long enough to draft a genuinely complete document, not a sketch of one.
Around the model, Anthropic shipped a Claude Code agent-teams research preview, an upgraded Claude in Excel, and a Claude in PowerPoint research preview for Max, Team, and Enterprise plans.
Pricing for Australians
At launch, Opus 4.6 cost US$5 per million input tokens and US$25 output in the API — roughly A$7.00 and A$35.00 at 1.40, billed in USD. The current flagship, Opus 5, carries the same US$5 / US$25 rate; the API is still priced in USD. App users get the flagship within existing Claude subscriptions, so the practical price of the upgrade is zero. Those subscriptions are now billed to Australians in AUD: Pro at A$28/month on annual billing (A$340 up front) or A$34 monthly, with Anthropic stating prices include 10% GST.
Australian Fit
Available in Australia from day one, and the writing remains the most natural of the major assistants in Australian English. Consumer billing has since moved to AUD with GST included. The rest of the structural picture is unchanged: US processing, no Australian data residency, no local support presence. For regulated work, the Privacy Act questions stay yours. The first-look 3/5 stands until a full re-score.
Early read
Good: the million-token window genuinely changes what you can hand it — a year of board papers, a full lease portfolio, an entire codebase; the effort dial is the most practical UI idea any frontier lab has shipped this year; Excel integration keeps improving. Not so good: 1M context was beta at launch (it is standard on the Claude 5 models) and long-context pricing on the API will sting at scale; agent teams are a research preview, not a product; and Anthropic still has no visible Australian footprint.
Early verdict
If you already pay for Claude, you just got a materially better tool for nothing — start feeding it bigger jobs. If you're choosing an assistant for long-document work — contracts, reports, compliance reviews — Opus 4.6 is now the benchmark the others have to beat.
| Criterion | Score |
|---|---|
| Value for Money | 4/5 |
| Ease of Use | 4/5 |
| Feature Depth | 5/5 |
| Australian Fit | 3/5 |
| Reliability & Support | 4/5 |
| Total | 20/25 ⭐⭐⭐⭐ Very Good |
First-look score — provisional until our full hands-on review.
Update history
- 9 Oct 2026 — Re-checked for the AUD-claim queue (#385): claude.com/pricing, read in a browser on an Australian connection, still shows Pro at A$28 a month on annual billing (A$340 up front) or A$34 monthly and Max from A$170.01, with "Prices include 10% GST" — the figures above stand. The API remains priced in USD. No claim changes. Capture:
Dossiers/captures/claude-pricing-2026-10-09.txt. - 20 Sep 2026 — Staleness check. Re-verified the model line-up, the API rates and Australian subscription pricing against the vendor. Line-up is now Fable 5.1 / Opus 5 / Sonnet 5 / Haiku 4.5, was Opus 4.8 (via 4.7) on our 9 July note. Opus 4.6 is a legacy model and no longer appears on the pricing page. The 1M-token context window is now standard on Fable 5.1, Opus 5 and Sonnet 5, was beta when Opus 4.6 introduced it. API rates for the current flagship Opus 5 are US$5 input / US$25 output per MTok — the same rate Opus 4.6 carried, unchanged. Australian subscription pricing is now native AUD, was USD-only billing. Pro is A$28/mo annual (A$340 up front) or A$34 monthly; Max starts at A$170.01/mo. The footer reads "Prices include 10% GST". Score unchanged at 20/25. Models · Pricing (claude.com/pricing as rendered to Australian visitors, read 19 Sep 2026) · Wayback: capture pending (archive.org unavailable 20 Sep 2026)
- 5 Sep 2026 — Staleness flag added at 211 days (house voice rule 8). No claims re-verified this pass; no score change.
- 9 Jul 2026 — Superseded-model note added. Opus 4.6 has since been succeeded by Opus 4.8 (via 4.7); forward-linked to the current Claude review. Original first look preserved unchanged as a historical record.
- 6 Feb 2026 — First reviewed.
Frequently asked questions
What is Claude Opus 4.6 and what's new in it?
Claude Opus 4.6 is Anthropic's new flagship model, released 5 February 2026. The headline is a 1-million-token context window (in beta) — roughly 700,000 words in one conversation. It adds adaptive thinking with four effort levels, context compaction for longer conversations, and output up to 128,000 tokens. We score it 20/25 — Very Good.
Is Claude Opus 4.6 available in Australia?
Yes, from day one. Australians get the flagship within existing Claude subscriptions at no extra cost, and those subscriptions are now billed in AUD with GST included. The rest is unchanged: US processing, no Australian data residency, and no local support presence. For regulated work, the Privacy Act questions stay yours. We score Australian Fit 3/5.
How much does Claude Opus 4.6 cost for Australians?
App users pay nothing extra — it arrives inside existing Claude subscriptions, which now cost Australians A$28/month on annual billing or A$34 monthly for Pro, GST included. In the API, Opus 4.6 launched at US$5 per million input tokens and US$25 output, and the current Opus 5 carries the same rate, still billed in USD. Long-context API pricing will sting at scale.
What are the effort levels in Claude Opus 4.6?
Adaptive thinking lets you choose low, medium, high, or max, trading speed against depth per task. A quick email summary doesn't need the same gear as a contract analysis, and now it doesn't cost the same time. It's the most practical UI idea any frontier lab has shipped this year.
Should Australians upgrade to Claude Opus 4.6?
Yes, if you already pay for Claude — you just got a materially better tool for nothing, so start feeding it bigger jobs. For long-document work — contracts, reports, compliance reviews — Opus 4.6 is now the benchmark the others have to beat. Our first-look score is 20/25, provisional until a full hands-on review.
Sources
- Anthropic — Introducing Claude Opus 4.6: https://www.anthropic.com/news/claude-opus-4-6 (primary) [Wayback snapshot TBC]
- Claude pricing: https://claude.com/pricing (primary) [Wayback snapshot TBC]
Price Watch · Free
Know before your AI bill changes
A short email when AI tools change what they charge Australians: the new price in AUD, what it means for you, and whether to switch. No hype, no spam, unsubscribe any time.