What's New

On October 26, 2026 (October 27 NZDT), OpenAI released GPT-6.1 Terra with the model ID gpt-6.1-terra. 1. Positioning: GPT-6.1 Terra upgrades GPT-5.6 Terra. It is fast and efficient, strong at computer use, visual reasoning, everyday coding and writing, with a much lower factual error rate; OpenAI says it completes computer-use tasks over three times faster than GPT-6 Astra and GPT-6.1 Sol. For the most complex coding tasks GPT-6.1 Sol or GPT-6 Astra remain the better choice. 2. Official price and PhotonMark 10% policy: OpenAI lists $1.00 / $0.05 / $5.00 per 1M input / cached input / output tokens, half the price of GPT-5.6 Terra and GPT-6.1 Sol. PhotonMark keeps deducting 10% of the official standard price: GPT-6.1 Terra is USD $0.10 / $0.005 / $0.50 per 1M input / cached input / output tokens. 3. Effective cash cost: with the current 70% top-up bonus (1.7x credit on USD $10+ top-ups, 1.734x with PayPlus), effective cash cost is about 5.9% of the official API price (roughly $0.059 input and $0.29 output per 1M tokens). 4. Billing rules: cache writes remain free; Fast mode keeps the established 2.5x credit multiplier; requests above 272K input tokens keep the same standard rates with no long-context surcharge (consistent with the GPT-6 family). Existing prices for gpt-5.6-terra, gpt-6.1-sol and other models are unchanged. 5. Availability: whether you can call the model depends on upstream account rollout, and appearing in the model list does not guarantee every account has access. Restart Codex to see the new model in the model list.

At DevDay on September 29, 2026 (September 30 NZST), OpenAI released GPT-6.1 Sol with the model ID gpt-6.1-sol and a 1,050,000-token context window. 1. Positioning: GPT-6.1 Sol upgrades GPT-6 Sol and comes close to GPT-6 Astra on agentic coding, computer use and professional work at one-fifth of Astra's standard API prices, with a lower factual error rate. OpenAI also said GPT-6.1 Astra will not be released because it did not meet its safety standards. 2. Official price and PhotonMark 10% policy: OpenAI lists $2.00 / $0.10 / $10.00 per 1M input / cached input / output tokens. PhotonMark keeps deducting 10% of the official standard price: GPT-6.1 Sol is USD $0.20 / $0.01 / $1.00 per 1M input / cached input / output tokens. 3. Effective cash cost: with the 70% top-up bonus in effect from October 20, 2026 (1.7x credit on USD $10+ top-ups, 1.734x with PayPlus), effective cash cost is about 5.9% of the official API price (roughly $0.12 input and $0.59 output per 1M tokens). 4. Billing rules: cache writes remain free; Fast mode keeps the established 2.5x credit multiplier; requests above 272K input tokens keep the same standard rates with no long-context surcharge (consistent with the GPT-6 family). Existing gpt-6-sol, gpt-6-luna and gpt-6-astra prices are unchanged. 5. Availability: whether you can call the model depends on upstream account rollout, and appearing in the model list does not guarantee every account has access. The Ultrafast speed tier for GPT-6.1 Sol / GPT-6 Astra is not currently supported by PhotonMark; pricing will be announced separately if it is added.

OpenAI officially released GPT-6 Sol and GPT-6 Luna on September 22, 2026, with model IDs gpt-6-sol and gpt-6-luna, featuring a 1,050,000-token context window. 1. Model Positioning: GPT-6 Sol is the all-round workhorse for complex coding, multistep verification, and agentic workflows. GPT-6 Luna is the lightweight, ultra-cost-efficient model for high-volume, everyday developer tasks. 2. Official Pricing & PhotonMark 10% Policy: OpenAI substantially lowered official API rates by over 50% compared to GPT-5.6 (Sol official is $2.00 / $10.00, Luna official is $0.10 / $0.50 per 1M tokens). PhotonMark strictly follows our 10% standard balance deduction policy: - GPT-6 Sol: USD $0.20 / $0.02 / $1.00 per 1M input / cached input / output tokens; - GPT-6 Luna: USD $0.01 / $0.001 / $0.05 per 1M input / cached input / output tokens. 3. Effective Cash Cost: With the 70% top-up bonus in effect from October 20, 2026 on USD $10+ top-ups (1.7x credit, or 1.734x on PayPlus), effective user cash cost is about 5.9% of official OpenAI API pricing (Sol effectively ~$0.12 / $0.59, Luna ~$0.006 / $0.029 per 1M tokens). 4. Billing Rules: Cache writes remain free ($0); Fast mode retains established 2.5x credit multiplier; single requests strictly exceeding 272K input tokens keep the same standard rates with no long-context surcharges (consistent with GPT-6 Astra and GPT-5.6).

PayPlus WeChat Pay and Alipay are now available to all users. Pay / Boost top-up amounts are USD $1, $2, $5, $10, $15, $20, $25, $30, $40 and $50, making it easier to pay for near-term needs. Every PayPlus amount receives the current top-up bonus followed by another 2% credit: from October 20, 2026 the bonus is 70%, for total credit of 1.734 times the base amount, so USD $10 credits $17.34. A 2% payment surcharge applies. Payments are collected in NZD; final RMB charges and exchange-rate offers are shown at checkout. New Stripe and PayPlus top-ups now have 30-day validity from payment. Unused balance is merged and an existing later expiry is preserved. Existing 60-day orders retain their agreed term. We encourage topping up for actual needs rather than keeping large prepaid balances. We do not promise permanent validity; we protect existing expiry dates and explain the rules clearly.

OpenAI announced GPT-6 Astra on September 4, 2026 NZST, with the model ID gpt-6-astra. Its official ChatGPT/Codex rate card now clarifies that Astra usage in Codex does not incur an additional long-context multiplier above 272K input tokens; Codex also does not charge for cache writes. PhotonMark has therefore changed gpt-6-astra to use its standard balance rate at every input length: USD $1 / $0.10 / $5 per 1M input / cached input / output tokens. Cache writes remain free, and Fast retains PhotonMark's established 2.5x credit multiplier. OpenAI is rolling out access gradually, so availability still depends on the upstream account and rollout.

Current Codex releases may enter Luna Reserve after the main ChatGPT usage window is exhausted and send requests under the internal model name gpt-reserve. PhotonMark has updated Codex Boost so every gpt-reserve request first uses the customer's own OpenAI authorization and Luna Reserve, regardless of the main-window status. A successful Luna Reserve request does not deduct Boost balance. Boost falls back to PhotonMark-managed GPT-5.6 Luna, billed at the Luna rate, only when the actual gpt-reserve request returns an explicit usage-limit response. The regular model choices shown by the Codex UI in Reserve mode are controlled by the Codex App / OpenAI, so Boost cannot add Sol or another option that the client does not offer; auto-review can still use the managed Boost service. This was a server-side update, and existing users do not need to change or regenerate their configuration.

PhotonMark AI Chat is now available as a beta. It is designed for users who do not want to install the Codex App, ChatGPT desktop client, or CLI: a web browser is all that is required to use supported PhotonMark Codex models. Sign in with your existing PhotonMark Cloud Account; no additional registration is required. Select PhotonMark Codex and a supported model such as GPT-5.6 Luna, Terra, or Sol to start chatting. Native web search is available without entering a Serper, Firecrawl, Jina, OpenAI, or other third-party API key; ask the model to search the web when you need current information. A successful search costs USD $0.001 under the existing Web Search rule, with ordinary model usage billed separately. AI Chat has no separate balance. Actual usage is deducted from your existing active Codex Pay balance and remains visible in the Codex Dashboard. Requests continue to follow PhotonMark Codex safety screening. This is a beta, so please contact support about login, model-selection, interrupted-response, or billing-display problems.

From July 31, 2026 at 13:05 NZST, automatic-review tasks recorded as codex-auto-review change from GPT-5.5 pricing to GPT-5.6 Luna pricing: USD $0.02 / $0.002 / $0.12 per 1M input / cached input / output tokens. Luna has no long-context surcharge, Cache write remains free, and requests explicitly marked Fast/Priority follow Luna's 2.5x rule. The new pricing applies only to usage incurred at or after the effective time. Already posted historical auto-review usage and deducted balances keep their original amounts and are not recalculated or adjusted.

PhotonMark now supports the standalone alpha/search endpoint used by the built-in Codex web.run tool across Codex Link, Codex Pay, and Codex Boost. Prefer the recommended Dashboard configuration with model_provider = "openai"; no local hotfix, custom model catalog, or manually injected tool is required. If an old [model_providers.*] block must be retained, its case-sensitive display name must be name = "OpenAI". If search still fails, remove local gateway rules that mutate alpha/search requests, restart Codex, and test again. Successful standalone searches follow the existing Web Search billing rule.

The referral reward continues: each time an existing PhotonMark Codex user successfully refers a new user who makes a single payment of at least US$10, the referrer can receive US$2 in Codex Pay or Codex Boost credit. There is currently no end date; we will provide separate notice when the offer ends. Every existing user now has a permanent referral link on the Referral tab of the Dashboard. The referral is bound automatically when a new user follows the link and first signs in to PhotonMark Codex, and remains recorded if they pay several days later. Existing Codex users, self-referrals, and referrals between accounts controlled by the same person are not eligible. After qualification, we manually review linked accounts, payment, refund, and dispute status before issuing the reward to the Pay or Boost destination selected in the Dashboard.

OpenAI officially released the GPT-5.6 family on July 10, 2026: Sol for frontier capability, Terra for a balance of intelligence and cost, and Luna for efficient, high-volume workloads. PhotonMark Codex supports all three explicit model names. Please refer to the pricing page for model rates, effective costs after top-ups, and the Fast/Priority and long-context billing rules.