Monthly roundup
July 2026 in AI models
A release-wave summary for readers who want the important changes without pretending every launch is a reason to switch.
The signal
17 releases plus 2 material status updates
DeepSeek, LG, Anthropic, InclusionAI, Moonshot AI, Google, Sakana AI, SpaceXAI, OpenAI, Meta account for 17 tracked release events. 16 catalogued events link to a detailed comparison profile. 6 launch guides are available for deeper trade-off coverage. There is no single winner because the constraints are different.
At a glance
- Release events
- 17
- Profile links
- 16
What shipped and changed
ReleaseDeepSeek-V4-Flash-0731
DeepSeek
Date note: The API keeps the `deepseek-v4-flash` model ID while serving the 0731 production update.
DeepSeek API changelog
ReleaseLG K-EXAONE 2.0
LG
LG AI Research announces the K-EXAONE 2.0 text family, a 750B/37B MoE under Apache 2.0; EXAONE 4.5 remains a separate 33B vision-language model.
Date note: The July 31 announcement links the BF16 checkpoint. The catalogue has a distinct FP8 profile with 262,144 native context; its exact publication day, hard output cap and hosted API tariff remain unverified.
LG AI Research news
- Status update
DeepSeek deepseek-chat / deepseek-reasoner retired
DeepSeek
DeepSeek fully retires the deepseek-chat and deepseek-reasoner aliases (15:59 UTC); R1 and V3.1 are no longer served by the first-party API.
DeepSeek — deepseek-chat / deepseek-reasoner retirement notice 
ReleaseLing 3.0 Flash
InclusionAI
Date note: Publisher article date July 24; initial hosted launch and later downloadable weights are separate stages. Native context is 262,144 positions. Launch free access ended August 3 at 23:00 Beijing time; no exact weights release day is verified.
Ant Ling Ling-3.0-flash release- Availability update
Ling 3.0 Flash native-platform launch
InclusionAI
Date note: Native chat/API availability. This date does not establish initial downloadable-weight publication or a later third-party hosting date.
Ant Ling official dated platform changelog 
Availability updateKimi K2.7 Code
Moonshot AI
Moonshot published the official Kimi K2.7 Code model page; the model itself was released on June 12.
Kimi K2.7 Code model page
Limited previewGemini 3.5 Flash Cyber
Google
Date note: Specialised model paired with Google's CodeMender security agent; access is not the same as a general public API release.
Google AI announcement
Availability updateSakana Fugu Cyber v1.0
Sakana AI
Date note: Security-focused orchestration API announcement. Current access requires manual approval and pay-as-you-go billing; the historical launch plan for Token Plan access is not current availability.
Sakana Fugu-Cyber launch
Availability updateGrok 4.5
SpaceXAI
SpaceXAI's formal Grok 4.5 launch post, a week after the July 8 API release.
SpaceXAI announcement
General availabilityGPT-5.6 Luna
OpenAI
General availability began after the limited preview.
OpenAI API changelog (Jul 9 2026)- General availability
GPT-5.6 Sol
OpenAI
General availability began after the limited preview.
OpenAI API changelog (Jul 9 2026) - General availability
GPT-5.6 Terra
OpenAI
General availability began after the limited preview.
OpenAI API changelog (Jul 9 2026) 
- Release
Grok 4.5
SpaceXAI
Date note: xAI's API release notes confirm the July 8 API release; the formal launch post followed on July 16.
xAI API release notes - Status update
Claude Fable 5 access
Anthropic
Global access was restored after the export controls were lifted.
Anthropic redeployment update
Read the important launches
OpenAI
GPT-5.6 Sol, Terra, and Luna: what the three-tier release changes
OpenAI's July family is a tier decision before it is a model decision: Sol for difficult work, Terra for balance, and Luna for volume. Since 2026-09-22 GPT-6 Sol and Luna are the newer Sol/Luna options at lower rates; input and output savings differ by tier.
Anthropic
Claude Opus 5: July 2026 launch, now superseded by Opus 5.5
At its July 2026 launch, Claude Opus 5 targeted complex agentic coding and long-horizon work with a 1-million-token context. Opus 5.5 launched on 2026-09-22 at lower $4/$20 rates; Anthropic now lists Opus 5 as Active (legacy), so compare Opus 5.5 first for new work.
Gemini 3.6 Flash: the efficient multimodal workhorse
At its July 21, 2026 launch, Gemini 3.6 Flash emphasized speed, multimodal input and lower output use. Google now labels it an older-generation model; compare the newer Flash release for current work.
Moonshot AI
Kimi K3: open weights, huge context, and a datacentre-sized trade-off
Kimi K3 brings frontier ambitions, multimodal input, and open weights — but its scale changes what self-hosting means.
DeepSeek
DeepSeek V4-Flash-0731: why the post-training update matters
Retired from the first-party API on 2026-09-10: DeepSeek says the legacy `deepseek-v4-flash` id temporarily routes to V4.1-Flash at its rates. This page preserves a historical note about the 0731 revision; it is not a current model recommendation.
xAI
Grok 4.5: xAI's July 2026 API model
This July 2026 guide records Grok 4.5's documented API surface; use xAI's current model and pricing pages to check the exact live limits and charges before deployment.