Fable 5 leaves the subscription today, and the meter starts running
As of today, Anthropic's Fable 5 is no longer included in Pro, Max, Team or Enterprise plans. Access now requires pre-purchased usage credits billed at API rates: $10 per million input tokens and $50 per million output, exactly double Opus 4.8 and the most expensive line on Anthropic's price list. The company says the move is temporary, driven by demand outrunning capacity after the model's redeployment, and that subscription access returns when capacity permits. For an operator the read is direct: if your workflows quietly standardized on Fable 5 over the past month, your marginal cost just went from flat to metered overnight. Audit which jobs actually need the top model and which run fine on Opus 4.8 at half the price. The labs have now shown they will reprice the frontier on short notice when supply tightens, so treat model choice as a line item, not a default.
Source: Original source
GPT-5.6 general access opens this week, with a three-tier price ladder
OpenAI's GPT-5.6 family hits its general access window between July 7 and 14, with pricing already confirmed: the flagship Sol at $5 per million input tokens and $30 output, mid-tier Terra at $2.50 and $15, and the small Luna at $1 and $6. Set that next to Anthropic's Fable 5 at $10 and $50 the same week and the pricing strategy divergence is the story. OpenAI is pricing its top model to be used everywhere; Anthropic is pricing its top model to ration scarce capacity. If you run production inference, this week is the right time to re-run your cost-per-task math across both stacks, because the gap between comparable tiers has rarely been this wide, and the vendor that wins your workload now tends to keep it.
Source: Original source
Gemini 3.5 Pro finally moves, two missed deadlines later
After missing its June commitment twice, Google's Gemini 3.5 Pro is now in expanded enterprise preview on Vertex AI and beginning a gradual rollout on the developer platform. The delay itself is the signal worth reading. Google does not usually slip public dates, and doing it twice in one quarter says the constraint is compute or reliability at scale, not marketing. That matches everything else visible this month: Anthropic rationing its top model, TSMC turning away orders, and every lab suddenly careful about who gets frontier access and when. If Gemini is in your stack, get into the preview now rather than waiting for general availability, because staged rollouts in a capacity-tight market have a way of staying staged longer than announced.
Source: Original source
OpenAI floats handing the US government a $42 billion stake
Reports over the weekend say Sam Altman has pitched giving the US government an equity stake in OpenAI valued around $42.6 billion, taking the idea directly to the President, the Commerce Secretary and the Treasury. This lands while the White House finalizes a voluntary frontier-model standards framework with OpenAI, Anthropic and Google, expected within days. Read the two together: the leading lab is trying to convert regulatory risk into partnership before the rules are written, and an ownership stake would make the US government financially invested in OpenAI winning. For everyone else in the market, a state-aligned OpenAI changes the competitive terrain on procurement, export policy and antitrust. Nothing is signed yet, but the fact that this offer exists tells you how the labs now rank political security against independence.
Source: Original source