Anthropic released Claude Opus 5.5, followed roughly an hour later by OpenAI’s GPT-6 Sol and GPT-6 Luna. Grok 4.7 and MiMo v2.6 launched the day before.
Prices have dropped across the board
GPT-6 Luna now costs half the price of its GPT-5.6 predecessor. GPT-6 Sol follows a similar reduction compared to the earlier model.
The current rates are as follows:
Current model pricing
- GPT-6 Luna: $0.10 per million input tokens, $0.01 cached input, $0.50 output
- GPT-5.6 Luna: $0.20 input, $0.02 cached, $1.20 output
- Grok 4.7: $2 input, $0.50 cached, $6 output
- GPT-6 Sol: $2 input, $0.20 cached, $10 output
- GPT-5.6 Terra: $2 input, $0.20 cached, $12 output
- Claude Opus 5.5: $4 input, $0.20 cached, $20 output
- GPT-5.6 Sol: $4 input, $0.40 cached, $20 output
- Claude Fable 5.1: $10 input, $0.25 cached, $50 output
- GPT-6 Astra: $10 input, $1 cached, $50 output
GPT-5.6 is scheduled for a 25% price increase in November. GPT-6 is priced at half the rate of those promotional figures.
With GPT-5.6 Terra now costing the same as GPT-6 Sol, the argument for using Terra has vanished.
Grok 4.7 priced itself at $2 and $6 per million tokens. This was less than half the price of GPT-5.6 Sol and matches GPT-6 Sol on input costs.
GPT-6 Luna is one of the cheapest models OpenAI has released. It beats GPT-4.1 Nano and GPT-5 Nano on price, though those older models are weaker.
I generated pelicans for GPT-6 Luna and GPT-6 Sol, then combined them with previous versions in a comparison grid. The 5.6 family used bolder, brighter colors. The 6 family is more muted. GPT-6 Astra on max produced the best result.
Opus 5.5 is cheaper and clearer
Opus 5.5 addresses complaints about the model’s communication style. Thariq Shihipar noted:
Opus 5.5 is the result of your feedback.
It communicates clearly, it’s cheaper per token than Opus 5.0 with the intelligence of Fable 5.1 it’s very token efficient and works across every effort level.
The model is also better at Blender. I plan to test it there.
Previous versions of Opus shared the same price: $5 per million input tokens and $25 per million output. Opus 5.5 is a 20% reduction to $4 and $20.
Cache read prices fell by 60%. This matters for long agentic conversations where 90% or more of input tokens are processed at cached rates.
The new price for Opus 5.5 matches the previous GPT-5.6 Sol price, before OpenAI cut their Sol rates in half.
GPT-6 Astra and Claude Fable 5.1 are both priced at $10 input and $50 output. The price war affects the tier below that.
Anthropic says Sonnet 5.5 and Haiku 5.5 are coming soon. It will be interesting to see if Haiku can regain price competitiveness at the lower end. Current Haiku 4.5 is $1 and $5 per million tokens, while GPT-6 Luna is one tenth of that at $0.10 and $0.50.
Opus 5.5 max over-thinks to the point of breaking
In a first for my test, Claude Opus 5.5 at “max” thinking level failed to return a response.
It started by calling this “a classic test request”, then thought really, really hard about what it was doing:
This is a classic test request, so I want to plan out a well-composed pelican with its distinctive beak and pouch riding a bicycle with proper wheels, frame, and pedals, set against a simple sky and ground backdrop. […]
Verifying the shin length checks out at roughly 95.2, close enough. Now I’m working out the near leg path from hip to knee to ankle, then sketching the foot shape resting on the pedal — outlining the heel, toe tips, and sole contour with a path using lines and curves to sit naturally on the pedal surface around y=478-494. […]
I like the fish sticking prominently out of the basket with the pelican eyeing it as a fun detail worth keeping. I’m also confirming the eye placement near the bill base matches typical pelican anatomy, and considering giving it a slightly happier expression. […]
The far leg reads correctly as passing behind the frame, so I’m moving on to check the chainring teeth and confirm layer ordering—the far crank arm should be mostly hidden by the seat tube and chainring. I’m settling on the final SVG’s width and height attributes alongside the viewBox to ensure proper scaling, noting there’s no text so no font-family is needed. […]
I was so excited to see this pelican… but then it stopped. Opus 5.5 has a 128,000 maximum output token limit, and it hit that while it was still reasoning about the SVG.
I tried a second time and got the same result. This makes me suspect that “max” is effectively useless. If it over-thinks to breaking point on a stupid SVG prompt, I do not trust it not to do the same for more interesting work.
(Those two failures each cost me $2.56 and took nearly 20 minutes.)
Fable 5.1 on “max” did not over-think and gave me the best pelican I have seen from any Anthropic model.
Here are the Opus 5.5 pelicans, excluding 5.5 max.
I also built a comparison grid comparing them with pelicans by Opus 5, Fable 5.1, and Sonnet 5.
Comparing different model vendors by how well they draw a pelican riding a bicycle may not make much sense now, but I am still finding value in using them for comparisons of the same model families at different reasoning levels.
I am now using GPT-6 Sol and Claude Opus 5.5 as my default models in Codex and Claude Code. I have upgraded the Datasette Agent demo to use GPT-6 Luna. It seems to be fast and competent at both SQL queries and building HTML and JavaScript for Datasette Apps.
