Pricing news

• Fast mode added for Claude Opus 5.5 on Claude in Microsoft Foundry for Azure-hosted deployments

• Claude Managed Agents example updated from Claude Opus 5 to Claude Opus 5.5
• Claude Managed Agents worked example costs decreased, including the cached example total

• Text editor tool billing changed: text_editor_20250429 was replaced by text_editor_20250728, with different token charges for Claude 4.7+ / Claude Mythos Preview vs Claude 4.6 and earlier.
• Bash tool billing rules were broadened: the higher token charge now applies across Claude 4.7+ models and Claude Mythos Preview, not just specific Opus models.
• Data residency billing clarified for Claude Haiku 5.5: the inference_geo: "us" surcharge also applies to its over-100k-token pricing tier.

• Claude Haiku 5.5 long-context billing rules were clarified: the 100K+ prompt threshold now explicitly includes cache reads and cache writes, and is applied per request, including cached prompt portions

• Claude Haiku 5.5 added, including batch pricing.
• Claude Haiku 5.5 introduced prompt-length tiered billing, with higher rates above 100,000 tokens; it is excluded from the standard long-context pricing rule.
• Claude Sonnet 5.5 cache-hit pricing decreased.
• Claude Haiku 5.5 tool-use billable token overhead added; Claude Haiku 3.5 retirement updated to remove Bedrock availability.

• Added Claude Sonnet 5.5, including standard and batch pricing
• Added billable tool-use system prompt token documentation for Claude Sonnet 5.5

• Claude Opus 5.5 added.
• Claude Opus 5.5 fast mode and batch pricing added.
• Claude Opus 5.5 prompt caching added, with cache reads/hits billed at a special lower multiplier than the standard cache-read rule.
• Claude Opus 5.5 tool-use billing/documentation added; tool choice support listed as auto and none only.

• Added Claude Fable 5.1 and Claude Mythos 5.1.
• Added pricing for Claude Fable 5.1 and Claude Mythos 5.1, including base input, cache writes, output, and cache hits/refreshes.
• Added a model-specific prompt-caching rule for Claude Fable 5.1 and Claude Mythos 5.1 with a reduced cache hit/refresh multiplier.
• Added batch pricing for Claude Fable 5.1 and Claude Mythos 5.1.

• Added billable pricing details for the new Browser use tool (browser_toolset_20260801), including token overhead rules and billable image/text tool-result usage.
• Computer use billing rules changed for the new computer_toolset_20260801, with new higher tool-definition overhead and zoom config affecting billable tokens.
• Prior computer use pricing figures are now explicitly limited to older tool versions (computer_20251124 and computer_20250124).
• Computer use billable input now explicitly includes zoom images, in addition to screenshots and tool execution results.

• Claude Sonnet 5 pricing increase was canceled; current pricing is now permanent
• Claude Sonnet 5 batch pricing increase was canceled; current batch pricing is now permanent

• Claude Opus 4.1 availability changed from deprecated to retired except on Bedrock and Google Cloud.
• Claude Managed Agents fast mode premium now applies.
• Claude Managed Agents data residency multiplier now applies for US inference geography.
• Claude Managed Agents non-applicable modifiers list was narrowed to Batch API discount and cloud platform pricing only.

• Gemini 4 Argon availability changed: it is no longer excluded from the Google AI Pro plan and is now only unavailable on the Free tier.

• Added a new invite-only Contributor pricing tier for Gemini 4 Argon, Gemini 3.8 Flash, Gemini 3.5 Flash-Lite, and Gemini 3.1 Pro Preview.
• Contributor tier introduces new billing/data-use rules: paid usage is used to improve products, and is documented as Standard-tier-equivalent for speed/reliability.
• Gemini 4 Argon and Gemini 3.8 Flash Contributor tier pricing has scheduled future price increases.
• Gemini 3.1 Pro Preview Contributor tier adds prompt-length-based pricing tiers; Gemini 3.5 Flash-Lite Contributor tier adds lower context-caching pricing than Batch/Flex.

• Added Gemini 4 Argon.
• Introduced Gemini 4 Argon pricing across Standard, Batch, Flex, and Priority.
• Gemini 4 Argon billing/availability added: not available on Free tier or Google AI Pro, implicit-only context caching, and paid Google Search/Maps grounding.
• Free-tier availability changed: Gemini 3.1 Pro Preview (Standard) changed to N/A; Gemini 3 Pro Image (Standard, Batch, Flex, Priority) changed to not available.

• Gemini 3.7 Flash removed
• Gemini 3.5 Flash removed

• Google Search grounding sharing scope changed from Gemini 3.x-only to all Gemini 3 and newer models.
• Google Maps grounding sharing scope changed from Gemini 3.x-only to all Gemini 3 and newer models.
• Google Maps billing wording changed from monthly free “prompts” to monthly free “requests” across affected Gemini 3.x model rows.

• Gemini Nano Banana 2.1 4K image pricing increased in both standard and batch tiers
• Gemini Nano Banana 2.1 4K image billing changed to consume more output tokens

• Added model: Gemini Nano Banana 2.1.
• Added pricing for Gemini Nano Banana 2.1 across Standard and Batch, including input, output, image output, and grounding/search usage.
• Added image-output billing rules for Gemini Nano Banana 2.1, including token consumption by image size.

• Removed: Gemini 2.5 Computer Use Preview
• Computer use tool billing rule updated to standard per-model token pricing
• Removed legacy billing reference to the Gemini 2.5 Computer Use Preview pricing table

• Gemini 3.8 Flash TTS added equivalent output pricing per 10 seconds of audio across Standard, Batch, Flex, and Priority tiers, with higher rates scheduled from Jan 1, 2027.
• Gemini 3.8 Flash-Lite TTS added equivalent output pricing per 10 seconds of audio across Standard, Batch, Flex, and Priority tiers, with higher rates scheduled from Jan 1, 2027.

• Added Gemini 3.8 Flash TTS and Gemini 3.8 Flash-Lite TTS
• Both new TTS models include Standard, Batch, Flex, and Priority pricing tiers
• Both new TTS models add context caching and cached-context storage billing, plus free tier availability for Standard input/output/context caching
• Both new TTS models have scheduled price increases starting Jan 1, 2027

• Gemini Robotics ER 2 Streaming Preview pricing was added, including paid input/output pricing and Google Search grounding billing with a free monthly allotment.
• Gemini 2.5 Computer Use Preview pricing changed from a time-based schedule with free-tier and search-grounding charges to prompt-length tiered input/output pricing.
• Gemini 2.5 Computer Use Preview no longer shows free-tier pricing or a separate Google Search grounding charge.

• Added gemini-3.8-live and gemini-3.8-live-extended-thinking.
• gemini-3.1-flash-live-preview pricing documentation was merged into a combined live section with those models.
• Added Google Search grounding billing rule for the combined live section, including a shared free tier and paid usage thereafter.

• Gemini 2.5 Flash Image was deprecated, with shutdown scheduled; migration is directed to Gemini 3.1 Flash Image or Gemini 3.1 Flash Lite Image.
• Gemini 2.5 Computer Use Preview gained a new pricing scheme, including a free tier, new paid input/output pricing, and Google Search grounding charges, while the older pricing block remains.
• Gemini Robotics ER 2 Streaming Preview pricing was removed/replaced, with its pricing content now appearing under Gemini 2.5 Computer Use Preview.
• File search billing now references Gemini Embedding 2 instead of Gemini Embedding, with no price change.

• Gemini Robotics ER 2 Preview pricing decreased temporarily through Dec 31, 2026 for standard and batch input/output, with prices reverting on Jan 1, 2027.
• Gemini Robotics ER 2 Preview context caching and cache storage pricing decreased temporarily through Dec 31, 2026 for both standard and batch, with prices reverting on Jan 1, 2027.
• Gemini Robotics ER 2 Streaming Preview standard input/output pricing decreased temporarily through Dec 31, 2026, with prices reverting on Jan 1, 2027.

• Lyria 3.5 preview models were consolidated into a single lyria-3.5 model
• lyria-3.5-clip-preview was removed
• Lyria 3.5 clip-tier pricing was removed; Lyria 3.5 is now priced as a single full-song model

• Added Lyria 3.5 model family: lyria-3.5-clip-preview and lyria-3.5-pro-preview
• Added pricing for Lyria 3.5 models
• Added billing/rate-limit note that Lyria 3.5 preview models have more restrictive rate limits

• Added Gemini 3.8 Flash.
• Gemini 3.8 Flash introduced Standard, Batch, Flex, and Priority pricing tiers.
• Gemini 3.8 Flash added billable Google Search grounding and Google Maps grounding rules.

• Added a billing-rule note for agentic video understanding: billing is based on content actually loaded rather than full video length, with variable token usage by query complexity and sampling depth.

• Added Gemini Omni Flash (gemini-omni-1.1-flash)
• Gemini Omni Flash is now generally available on the paid tier
• Gemini Omni Flash pricing was added
• Gemini Omni Flash Preview remains alongside the new GA model

• Removed Gemini 2.5 Flash-Lite Preview, Gemini 2.0 Flash, and Gemini 2.0 Flash-Lite from the pricing page
• Removed Imagen 4, Veo 3, and Veo 2 from the pricing page

• Added Gemini 3.5 Transcribe Live with standard input/output pricing
• Added Gemini 3.5 Transcribe with standard input/output pricing
• Gemini 3.5 Transcribe and Gemini 3.5 Transcribe Live do not support Google Search grounding
• Added billing-rule notes for Gemini 3.5 Transcribe and Gemini 3.5 Transcribe Live using estimated audio/text token conversion rates

• Added Gemini 3.7 Flash with Standard, Batch, Flex, and Priority pricing plus grounding pricing.
• Gemini 3.6 Flash got a temporary 50% discount through Dec 31, 2026 across Standard, Batch, Flex, and Priority.
• Gemini 3.6 Flash context caching and context caching storage pricing were also temporarily reduced through Dec 31, 2026, then revert Jan 1, 2027.

• Added Life sciences models section with gpt-rosalind-research and gpt-rosalind-discovery.
• gpt-rosalind-research was removed from Specialized models and its billing/access note moved to Life sciences models.

• gpt-6.1-sol was added to Ultrafast pricing, including short- and long-context tiers.

• Added Batch-tier pricing for gpt-image-2.5-sunburst and gpt-image-2.5-flare.

• Added gpt-6.1-sol, with lower cached-input pricing than gpt-6-sol across Standard, Batch, Flex, and Fast; other pricing matches gpt-6-sol.
• Added Ultrafast pricing tier for gpt-6-astra.
• Uplift/billing rules changed: regional processing and FedRAMP are now stated as 10% uplifts; billing-through-platform notes now explicitly include Microsoft Azure, and the Bedrock direct-price equivalence note was removed.
• Removed alias billing note for gpt-daybreak-blue-latest/gpt-daybreak-red-latest and removed stated data-residency availability constraints from flagship pricing sections.

• Cached input billing for GPT Image 2 and GPT Image 2.5 now applies only to images generated with the Responses API.

• Added a 10% surcharge for FedRAMP endpoints relative to corresponding standard model rates.

• Added new flagship models: gpt-6-sol and gpt-6-luna.
• Added pricing for gpt-6-sol and gpt-6-luna across Standard, Batch, Flex, and Fast modes.
• Updated EU data residency/billing rules for GPT-6 models: gpt-6-sol and gpt-6-luna support EU data residency only with Standard processing.
• Fast mode EU data residency rule changed for gpt-6-astra and now also applies to gpt-6-sol and gpt-6-luna.

• gpt-5.4-cyber was removed from the Cyber models pricing table

• Added gpt-rosalind-research under Life Sciences
• gpt-rosalind-research billing starts October 5, 2026, with no cache-write pricing
• Added a 10% regional processing surcharge for eligible data residency endpoints, including Specialized models / Fast mode

• Added GPT-Live sessions billing.
• Added gpt-live-1.
• GPT-Live voice sessions are billed per second with no rounding to whole minutes.
• GPT-Live backend model and tool usage is billed separately.

• Amazon Bedrock billing rule updated: commercial region pricing now matches direct OpenAI pricing for equivalent services.

• Added Standard pricing for image generation models gpt-image-2.5-sunburst and gpt-image-2.5-flare

• Added new model: gpt-6-astra
• Added gpt-6-astra pricing for Standard, Batch, Flex, and Fast mode
• Added billing/availability rule: Fast mode is unavailable for gpt-6-astra with EU data residency; Standard must be used instead

• Cyber model aliases renamed: daybreak-blue-latest → gpt-daybreak-blue-latest; daybreak-red-latest → gpt-daybreak-red-latest

• gpt-5.6-sol prices decreased across Standard, Batch, Flex, and Fast mode, including short- and long-context rates
• Cyber models pricing for gpt-5.6-sol decreased to match the updated base-model rates
• Added promotional billing note for gpt-5.6-sol pricing availability through November 21, 2026

• Added gpt-5.6-cyber.
• Added aliases daybreak-blue-latest → gpt-5.6-sol and daybreak-red-latest → gpt-5.6-cyber.
• Removed gpt-5.3-chat-latest and gpt-5.2-chat-latest.

• Added long-context pricing for gpt-5.6-sol, gpt-5.6-terra, and gpt-5.6-luna in Fast mode.

• Voice model changes: Speech to Speech (grok-voice-think-fast-2.0) removed/renamed; Speech to Speech (grok-voice-3.0) added
• Deprecated model entry added: Speech to Speech (grok-voice-2.0) marked deprecated

• Added grok-imagine-video-1.5-lite to Imagine pricing

• File Attachments tool (attachment_search) price decreased.
• Grok 4.7 Fast long-context billing changed to a lower multiplier; standard token prices unchanged.

• X Search billing changed from per-call pricing to per-item billing.
• X Search now bills separately for posts fetched and user profiles fetched.
• X Search billing rules now count all returned posts in searches/thread fetches, including parent and quoted posts, and all profiles returned in user searches.

• Added grok-4.7 to Text API pricing.
• Added grok-4.7 Fast pricing for Cursor and Grok Build only, with long-context billing above 200k prompt tokens.
• US Regional Endpoint now includes grok-4.7 alongside grok-4.6.
• Added US Regional Endpoint pricing for grok-4.7.

• Added grok-imagine-image-quality and grok-imagine-image-2.0 image models with pricing.

• Removed Grok imagine image models: grok-imagine-image-quality and grok-imagine-image-2.0

• Added image models: grok-imagine-image-2.0 and grok-imagine-image-quality

• Removed Grok imagine image models: grok-imagine-image-2.0 and grok-imagine-image-quality

• Added US regional endpoint billing with a surcharge on token rates, including long-context and cached input tokens
• Prompt caching discounts are now applied before the US regional endpoint surcharge
• US regional endpoint availability added for grok-4.6 only

• Removed deprecated Speech to Speech (grok-voice-think-fast-1.0) from voice pricing

• X Search billing will change to usage-based billing by posts fetched and user profiles fetched
• X Search billing rules updated: all returned posts, including parent and quoted posts, and all profiles returned by user search are billable

Initial Commit