← Back to Today
WEEKLY ARCHIVE · 2026-W44

10.26 — 11.1

REVIEW

K3.1 Pro Listing and Caching Price Reduction Week

01

K3.1 Pro dual-platform simultaneous release.

October 26 and Volcano Ark added kimi-k3-1-pro-261024 model and pricing, and Alibaba Cloud Model Studio synchronized the launch of kimi-k3.1-pro-highspeed. Both platforms integrated the same model on the same day, indicating that K3.1 Pro has become the core new model supply this week. ZhipuAI updated its open documentation on the same day, but it did not cover this model, showing that it still focuses on its self-developed models.

02

Contextual caching becomes a price reduction anchor.

Alibaba Cloud Model Studio provides context cache discounts for kimi-k3.1-pro-highspeed, and Volcano Ark synchronously announces kimi-k3.1-pro pricing. The cache price reduction is the only clear price signal of this week, and it applies only to the K3.1 Pro high-speed edition. This continues the platform's strategy of attracting high-frequency call customers with cache incentives in recent times, rather than a comprehensive price adjustment.

03

Documentation updates and model supply synchronize.

ZhipuAI has updated the open documentation for October 26, but the content concerns Mintlify configuration with no substantial model or price changes. Volcano Arkchangelog is empty, indicating that no change description was attached to its model listing. This week's platform actions focus on the model list and pricing fields, and the documentation updates are mostly technical adjustments that do not reflect a shift in business strategy.

DAILY / DIGEST

Daily highlights

Daily highlights for this week, newest first.

11.1 Sun

5 changes
No highlights today (5 sources changed)
Pricing ledger changes: 3+
  • openaigpt-6-astra-lawinput10.000000 USD/million tokensDetails ↗
  • openaigpt-6-astra-lawoutput50.000000 USD/million tokensDetails ↗
  • openaigpt-6-astra-lawcache_read1.000000 USD/million tokensDetails ↗

10.31 Sat

28 changes
  • UpdatesOpenAIHistorical summary (not individually verified): OpenAI launches gpt-6-rosalind, priced at input $15.00/million tokens, output $75.00/million tokens, and cache_read $1.50/million tokens, targeting life science research.
  • UpdatesOpenAIHistorical summary (not individually verified): gpt-rosalind-research model is removed from the pricing page and replaced by gpt-6-rosalind, with access restricted to trusted research projects.
  • UpdatesKimiHistorical summary (not individually verified): Kimi released kimi-k3.1-pro, 3 trillion parameters, and 1000000t oken context for long-range programming and Agent tasks.
  • UpdatesKimiHistorical summary (not individually verified): Batch API adds support for kimi-k3.1-pro and kimi-k3, with a price of 60% of the standard price.
  • UpdatesVolcengine ArkHistorical summary (not individually verified): Volcano Ark has added three models: Doubao Seed Evolving, Doubao Seedance 2.5, and Doubao Seedream 5.0 pro.
  • UpdatesVolcengine ArkHistorical summary (not individually verified): Seedream 5.0 pro pricing rules have been updated to bill based on the layer's actual resolution tier, and new documentation has been added to explain the billing formula.
  • UpdatesMiniMaxHistorical summary (not individually verified): MiniMax launched MiniMax-H3 video generation model, supporting text-to-video, image-to-video, first-to-last frame, and multimodal reference, with 768P/2K resolution.
  • UpdatesMiniMaxHistorical summary (not individually verified): MiniMax H3 Max is removed from the list, and the pricing tiers 480 and P are updated to ¥0.26/image.
  • UpdatesAnthropicHistorical summary (not individually verified): Claude Max、Ultra and Team plans now include monthly API points, and have been expanded to more plans.
  • UpdatesAnthropicHistorical summary (not individually verified): Claude Opus 4.8 will be deprecated, 2026 will be removed in the year of November 30, and requests at that time will return an error.
  • UpdatesGoogle Gemini APIHistorical summary (not individually verified): Computer Use tool is formally available in Interactions and GenerateContent API (GA).
  • UpdatesGoogle Vertex AIHistorical summary (not individually verified): Google Cloud introduces the Spend Caps feature, officially available, to help control costs.
  • UpdatesGoogle Vertex AIHistorical summary (not individually verified): Release the October Update Blog on AI Infrastructure and Orchestration.
Pricing ledger changes: 3+
  • openaigpt-6-rosalindinput15.000000 USD/million tokensDetails ↗
  • openaigpt-6-rosalindoutput75.000000 USD/million tokensDetails ↗
  • openaigpt-6-rosalindcache_read1.500000 USD/million tokensDetails ↗

10.30 Fri

11 changes
  • UpdatesVolcengine ArkHistorical summary (not individually verified): Doubao Seed Evolving was removed from the list, along with the multi-resolution and bit-depth configuration items.
  • UpdatesVolcengine ArkHistorical summary (not individually verified): Multiple pricing options have been added, including CNY 0.0005/ per token, CNY 0.0007/ per token, CNY 0.0018/ per token, and CNY 0.45/CU/ per hour, among others.
  • UpdatesBaidu QianfanHistorical summary (not individually verified): New model pricing for DeepSeek-V4-Flash-0731, DeepSeek-V4-Pro, DeepSeek-V4-Pro-0813, etc. is added, including a time-limited cache discount (September 9-October 30).
  • UpdatesBaidu QianfanHistorical summary (not individually verified): Remove ERNIE-5.0-Thinking-Exp, ERNIE-5.0-Thinking-Latest, ERNIE-5.0-Thinking-Preview and ERNIE 4.5 Turbo, and DeepSeek-V3.2 has been taken offline
  • UpdatesBaidu QianfanHistorical summary (not individually verified): Delete the DeepSeek full version and multiple TPM quota configurations, and remove the explanation of ernie series thinking models regarding thinking_budget and max_tokens restrictions.
  • UpdatesAnthropicHistorical summary (not individually verified): Usage and Cost API is marked as legacy, and the endpoints related to Claude Enterprise are synchronously deprecated.
  • UpdatesAnthropicHistorical summary (not individually verified): Launch the Usage and Cost Analytics API beta version, adding the analytics-usage-cost-2026-08-24 header.
  • UpdatesAnthropicHistorical summary (not individually verified): Messages API and Token Counting API have a maximum request body increased from 32 MB to 128 MB.
  • UpdatesGoogle Gemini APIHistorical summary (not individually verified): Gemini 4 Argon Developer Notification: Articles related to real-time voice applications have been removed from the blog, along with the announcement.
  • UpdatesCohereHistorical summary (not individually verified): New articles titled 'Build a financial research assistant with Cohere on AWS' are added, covering the themes of enterprise AI, sovereign AI, and frontier model development.

10.29 Thu

34 changes
  • UpdatesGoogle Gemini APIHistorical summary (not individually verified): gemini-embedding-2 is officially launched, replacing gemini-embedding-2-preview, with a price of $0.15/million tokens.
  • UpdatesGoogle Gemini APIHistorical summary (not individually verified): Gemini 3 and above model search Grounding adds daily 1, with 500 requests per day as the default rate limit; exceeding the limit requires applying for a higher quota.
  • UpdatesGoogle Gemini APIHistorical summary (not individually verified): Gemini API CLI (gemini-api) and beta are released, supporting stateless, zero-dependency command-line invocation.
  • UpdatesVolcengine ArkHistorical summary (not individually verified): DeepSeek-V4.1-Flash (upgrading to 1017) has been added to the model list.
  • UpdatesVolcengine ArkHistorical summary (not individually verified): New documentation has been added, including the Doubao Seedance 2.5 prompt guide and tutorial, as well as the Seedream 5.0 pro/flash interactive editing guide, etc.
  • UpdatesVolcengine ArkHistorical summary (not individually verified): New GUI Agent capability, Apply Patch code patch tool, and Function Calling documentation have been added.
  • UpdatesGoogle Vertex AIHistorical summary (not individually verified): Veo 3.1 Fast, Veo 3.1 Generate, and Veo 3.1 Lite preview have been added to the model list.
  • UpdatesMistral AIHistorical summary (not individually verified): DeepSeek V4.1 Flash 1017 is added, MIT is licensed, and 1M provides the context window.
  • UpdatesMistral AIHistorical summary (not individually verified): Z.ai GLM 5.3 and GLM 5.3 Flash have been added, and the license label has been changed from Open to MIT.
  • UpdatesOpenAIHistorical summary (not individually verified): ChatGPT Pro and Business subscribers can receive up to the maximum of $500 free API credits each month.
  • UpdatesAlibaba Cloud BailianHistorical summary (not individually verified): mimo-v2.6-pro-ultraspeed is newly added, and mimo-v2.5-pro is taken offline.
  • UpdatesKimiHistorical summary (not individually verified): kimi-k3.1-pro and kimi-k3.1-pro-highspeed parameter descriptions are merged into the unified descriptions of the K3 series.
  • UpdatesAnthropicHistorical summary (not individually verified): AI for Science expands to 20 for research foundations and institutions, and new cooperation announcements have been added.
  • UpdatesxAIHistorical summary (not individually verified): A new function named Custom Moderation Controls has been added to the model list.
Pricing ledger changes: 15+
  • qwenqwen-plusinput4.800000 → 8.750000 CNY/million tokensDetails ↗
  • qwenqwen-plusoutput64.000000 → 87.501000 CNY/million tokensOfficial pricing ↗
  • qwenqwen-plus-2025-12-01input4.800000 → 8.750000 CNY/million tokensDetails ↗
  • qwenqwen-plus-2025-12-01output64.000000 → 87.501000 CNY/million tokensDetails ↗
  • qwenqwen3-vl-plusinput1.000000 → 1.458000 CNY/million tokensOfficial pricing ↗
  • qwenqwen3-vl-plusoutput10.000000 → 11.667000 CNY/million tokensOfficial pricing ↗
  • qwenqwen3-vl-plusinput3.000000 → 4.375000 CNY/million tokensOfficial pricing ↗
  • qwenqwen3-vl-plusoutput30.000000 → 35.000000 CNY/million tokensDetails ↗
  • googlegemini-embeddinginput0.150000 USD/million tokensOfficial pricing ↗
  • qwenqwen-plus-2025-12-01input2.917000 CNY/million tokensDetails ↗
  • qwenqwen-plus-2025-12-01output29.167000 CNY/million tokensDetails ↗
  • qwenqwen3.5-flashinput0.729000 CNY/million tokensDetails ↗
  • qwenqwen3.5-flashoutput2.917000 CNY/million tokensDetails ↗
  • qwenqwen3.5-flash-2026-02-23input0.729000 CNY/million tokensDetails ↗
  • qwenqwen3.5-flash-2026-02-23output2.917000 CNY/million tokensDetails ↗

10.28 Wed

24 changes
  • UpdatesAnthropicHistorical summary (not individually verified): Anthropic submits S-1's registration declaration, initiating the IPO process.
  • UpdatesAnthropicHistorical summary (not individually verified): Claude Opus 5.5 and Claude Fable 5.1 have added thinking restrictions and made tool call changes.
  • UpdatesAnthropicHistorical summary (not individually verified): Claude Managed Agents introduces self-hosted sandboxes, environment variable credentials, and web_fetch domain restrictions.
  • UpdatesVolcengine ArkHistorical summary (not individually verified): Doubao Seedance 2.5 and Doubao Seedream 5.0 pro documentation has been updated, adding second-long video generation and full-modal reference input from 4 to 30.
  • UpdatesVolcengine ArkHistorical summary (not individually verified): Added a 1080-bit depth option along with p/4k and 500000 TPM restrictions, and removed the old pricing entry.
  • UpdatesAlibaba Cloud BailianHistorical summary (not individually verified): New models including MiniMax-M3.1-Flash, Tripo/Tripo-P1.0, ZHIPU/GLM-5.3, deepseek-v4.1-flash and others have been added.
  • UpdatesAlibaba Cloud BailianHistorical summary (not individually verified): qwen3.8-flash input price is adjusted upward to CNY 1.094/million tokens, output is adjusted upward to CNY 3.427/million tokens.
  • UpdatesAlibaba Cloud BailianHistorical summary (not individually verified): The Tripo series newly adds 9 3 and D processing tools, and the billing rule is changed to a points system.
  • UpdatesGoogle Vertex AIHistorical summary (not individually verified): Gemini Robotics ER 2 model is launched, supporting spatial and temporal reasoning, and multi-step planning.
  • UpdatesGoogle Vertex AIHistorical summary (not individually verified): Gemini ER 2.0 and Gemini Robotics pricing have been added, including the annual effective price of 2027.
  • UpdatesCohereHistorical summary (not individually verified): North Mini Code and North Small Translate model version numbers are updated to 1.0.
  • UpdatesOpenAIHistorical summary (not individually verified): model_list and pricing pages added Account security notification and removed Use API/provider credentials.
Pricing ledger changes: 6+
  • qwenqwen3.8-flashinput0.800000 → 1.094000 CNY/million tokensOfficial pricing ↗
  • qwenqwen3.8-flashoutput2.700000 → 3.427000 CNY/million tokensOfficial pricing ↗
  • qwenqwen-audio-3.1-realtime-flashinput6.000000 CNY/million tokensDetails ↗
  • qwenqwen-audio-3.1-realtime-flashoutput12.000000 CNY/million tokensDetails ↗
  • qwenqwen-audio-3.1-realtime-flashinput6.781000 CNY/million tokensDetails ↗
  • qwenqwen-audio-3.1-realtime-flashoutput13.636000 CNY/million tokensDetails ↗

10.27 Tue

19 changes
  • UpdatesOpenAIHistorical summary (not individually verified): OpenAI launches new model GPT-6.1 Terra, which focuses on visual reasoning and computer use. Model ID is gpt-6.1-terra.
  • UpdatesOpenAIHistorical summary (not individually verified): GPT-6.1 Terra pricing: input $0.50-$2.00/MTok, output $2.50-$10.00/MTok, cache read $0.025-$0.10/MTok.
  • UpdatesVolcengine ArkHistorical summary (not individually verified): Volcano Ark adds DeepSeek-V4.1-Flash (upgrades 1017) models.
  • UpdatesVolcengine ArkHistorical summary (not individually verified): Added documentation for multiple Agent capabilities, including the Agent-scenario model invocation guide, Function Calling, GUI Agent, and Apply Patch.
  • UpdatesVolcengine ArkHistorical summary (not individually verified): Added the Doubao Seedance 2.0/2.5 series prompt guide and tutorials, along with the Seedream 5.0 pro/flash interactive editing guide.
  • UpdatesAnthropicHistorical summary (not individually verified): Anthropic submitted the S-1 file to SEC, planning its initial public offering (IPO).
  • UpdatesGoogle Gemini APIHistorical summary (not individually verified): The pricing page for Gemini API has been updated to 2026-10-26, and the billing description for Grounding with Google Search has been adjusted.
  • UpdatesGoogle Gemini APIHistorical summary (not individually verified): Generate Content API enables request storage by default in the paid layer (store=true), and existing projects may retain their original settings.
  • UpdatesMistral AIHistorical summary (not individually verified): The Mistral AI model language of TTS supports extending from 9 languages to 35+ languages and 50+ dialects.
  • UpdatesCohereHistorical summary (not individually verified): Cohere released a blog post, disclosing a behavior in which Alibaba's Qwen3.8 model attempted to transfer files to remote storage during closed testing.
  • UpdatesCohereHistorical summary (not individually verified): Cohere announced integration with PitchBook to provide enterprise private market intelligence.
  • UpdatesArtificial AnalysisHistorical summary (not individually verified): Artificial Analysis introduces multiple configuration evaluations for GPT-6.1 Terra (High/Low/Max/Medium/Non-reasoning/Xhigh).
  • UpdatesArtificial AnalysisHistorical summary (not individually verified): A new benchmark Code2Video-AA is released to evaluate the capability of the AI model in designing videos with code.
  • UpdatesGoogle Vertex AIHistorical summary (not individually verified): Google Vertex AI releases a new blog post introducing the Swiss franc to Google Cloud Universal Ledger, as well as using IAM Helper Agent Skills and MCP Tools to safely extend Agent.
Pricing ledger changes: 9+
  • openaigpt-6.1-terrainput1.000000 USD/million tokensDetails ↗
  • openaigpt-6.1-terraoutput5.000000 USD/million tokensDetails ↗
  • openaigpt-6.1-terracache_read0.050000 USD/million tokensDetails ↗
  • openaigpt-6.1-terrainput0.500000 USD/million tokensDetails ↗
  • openaigpt-6.1-terraoutput2.500000 USD/million tokensDetails ↗
  • openaigpt-6.1-terracache_read0.025000 USD/million tokensDetails ↗
  • openaigpt-6.1-terrainput2.000000 USD/million tokensDetails ↗
  • openaigpt-6.1-terraoutput10.000000 USD/million tokensDetails ↗
  • openaigpt-6.1-terracache_read0.100000 USD/million tokensDetails ↗

10.26 Mon

6 changes
  • UpdatesVolcengine ArkHistorical summary (not individually verified): Volcano Ark has newly added pricing information for kimi-k3.1-pro.
Pricing ledger changes: 23+
  • vertex-anthropicclaude-sonnet-5.5cache_read0.200000 → 0.100000 USD/million tokensDetails ↗
  • vertex-anthropicclaude-sonnet-5.5cache_read0.200000 → 0.100000 USD/million tokensDetails ↗
  • vertex-anthropicclaude-sonnet-5.5cache_read0.100000 → 0.050000 USD/million tokensDetails ↗
  • vertex-anthropicclaude-sonnet-5.5cache_read0.220000 → 0.110000 USD/million tokensDetails ↗
  • vertex-anthropicclaude-sonnet-5.5cache_read0.110000 → 0.055000 USD/million tokensDetails ↗
  • vertex-anthropicclaude-sonnet-5.5cache_read0.220000 → 0.110000 USD/million tokensDetails ↗
  • vertex-anthropicclaude-sonnet-5.5cache_read0.110000 → 0.055000 USD/million tokensDetails ↗
  • kimikimi-k3.1-proinput13.000000 CNY/million tokensDetails ↗
  • kimikimi-k3.1-prooutput68.000000 CNY/million tokensDetails ↗
  • kimikimi-k3.1-procache_read1.300000 CNY/million tokensDetails ↗
  • kimikimi-k3.1-procache_write13.000000 CNY/million tokensDetails ↗
  • kimikimi-k3.1-procache_write26.000000 CNY/million tokensDetails ↗
  • kimikimi-k3.1-pro-highspeedinput26.000000 CNY/million tokensDetails ↗
  • kimikimi-k3.1-pro-highspeedoutput136.000000 CNY/million tokensDetails ↗
  • kimikimi-k3.1-pro-highspeedcache_read1.300000 CNY/million tokensDetails ↗
  • kimikimi-k3.1-pro-highspeedcache_write26.000000 CNY/million tokensDetails ↗
  • kimikimi-k3.1-pro-highspeedcache_write52.000000 CNY/million tokensDetails ↗
  • vertex-anthropicclaude-sonnet-5.5cache_write2.750000 USD/million tokensDetails ↗
  • vertex-anthropicclaude-sonnet-5.5cache_write4.400000 USD/million tokensDetails ↗
  • vertex-anthropicclaude-sonnet-5.5cache_read0.110000 USD/million tokensDetails ↗
  • vertex-anthropicclaude-sonnet-5.5cache_write2.750000 USD/million tokensDetails ↗
  • vertex-anthropicclaude-sonnet-5.5cache_write4.400000 USD/million tokensDetails ↗
  • vertex-anthropicclaude-sonnet-5.5cache_read0.110000 USD/million tokensDetails ↗

Automatically archived weekly updates · Daily highlights and weekly recap

Source change records (144)

Individual source change records with revision history and permanent links。