OpenAI and Anthropic have cut prices on their flagship AI models by as much as 80% over the past four months, as cheaper open-weight rivals intensify competition and both companies face questions about possible public listings.
The reductions, which began in June, have lowered API usage costs by 75% or more in some cases. Developers in Pakistan who build with ChatGPT, Claude, Codex and Claude Code will see the biggest benefit on long coding sessions that repeatedly read cached conversation history.
What changed on the API price sheet
OpenAI began the latest round in late June when it launched GPT-5.6 Luna at $1 per million input tokens and $6 per million output tokens. By 30 July it had cut both prices by 80% to $0.20 and $1.20.
Anthropic followed in August by making a temporary launch discount for Claude Sonnet 5 permanent, then in September reduced Claude Opus 5.5 pricing by 40% to 60%. Days later OpenAI introduced GPT-6.1 Sol at half the input and output prices of its predecessor and cut cached-input pricing to 5% of the regular input cost, matching Anthropic.
For AI coding tools, cache reads can account for 96% to 99% of token spending, so a 75% cut in cache-read prices can reduce effective usage costs by about 70% to 75%, according to pricing trackers cited by ProPakistani.
Open-weight rivals keep pressure on
DeepSeek and Moonshot AI have increased pressure by offering lower-cost models through platforms such as OpenRouter. DeepSeek’s Flash V4.1 1017, released in April, lists input, output and cached token prices of $0.15, $0.60 and $0.003 per million tokens and is available through dozens of inference providers.
Moonshot AI’s Kimi K3.1 Pro is reported to deliver performance comparable to Claude Opus 5.5 at about half the input and output price, with weights planned for release on 4 November. Despite lower sticker prices, open-weight models still account for a minority of total token usage.
Why it matters for freelancers and startups here
Cheaper flagship APIs matter for Pakistani freelancers, agencies and startups that bill clients in dollars but pay for tokens in the same currency. Lower unit costs can stretch ChatGPT and Claude subscriptions and coding-agent quotas, though OpenAI also cut the usage allowance on its $200-per-month Codex tier by 50%, citing more efficient models.
Both companies are also linked to IPO talk: OpenAI is rumoured to target a 2027 listing, while Anthropic could file within months. Price cuts may squeeze per-token margins even as agent usage rises. Data from Vercel’s AI model gateway shows average cost per token has fallen to about 43% of its April 2026 level, while total spending still climbs because better models drive more use.
