AI Release RadarxAI12 Aug 2026
Grok 4.6
Keeps xAI’s rapid frontier-model cadence visible while full model-card, benchmark and pricing details remain to be verified.
Source is social-firstPricing not verifiedContext not verifiedAvailability unclear
Frontier model brief · Checked several times daily
The latest moves
in machine intelligence.
A signal-first view of the newest models from the labs shaping AI—what launched, what changed, and why it matters.
Refresh pulse · today27 sources checked
New launches0
Briefs upgraded0
Review cleared0
Source issues0
No major launch signal is still a result: the refresh ran and the index stayed conservative.
Editorial check queue
Needs review.
These launches are visible on the radar, but at least one important field still needs stronger first-party confirmation before treating it as fully verified.
Meta09 Jul 2026
Muse Spark 1.1
Marks Meta’s return to frontier APIs with native multi-agent orchestration and active context management.
Pricing not verified
Mistral AI08 Jul 2026
Robostral Navigate
Shows a compact model moving beyond screens into autonomous navigation across real environments.
Pricing not verifiedContext not verified
Verified release stream
10 labs. 13 verified launches.
Only first-party announcements. No composite scores, rumours or leaderboard theatre.
13 releasesAll capabilities
0113 Aug 2026
Gemini 3.7 Flash
ImpactMajor
High-signal launch with broad capability, access or platform-level implications.
Google’s latest Flash model is generally available in the Gemini API with improved multimodal and coding performance.
ContextunknownAccessAPIWeightsclosedPricevisibleCheckConfirmed
Availability
StatusGAChannelAPIWeightsClosedPricingKnownDocsLinked
Source qualityFirst-party model card
Primary model documentation is available and linked from this brief.
Strong source · 2 receiptsWhat changed
Model launchpricing visibleAPI access
Why it matters
Moves the Flash tier forward from 3.6 to 3.7 and gives developers a newer low-latency default with an introductory price window.
GeneralCodingMultimodal
0212 Aug 2026
Grok 4.6
ImpactNeeds review
Potentially important, but source strength or availability still needs editorial confirmation.
xAI’s latest Grok model release, announced by SpaceXAI on X as the successor to Grok 4.5.
ContextunknownAccessAPIWeightsclosedPriceunknownCheckreview
Availability
StatusNeeds checkChannelAPIWeightsClosedPricingUnclearDocsPending
What changed
Model launchAPI access
Why it matters
Keeps xAI’s rapid frontier-model cadence visible while full model-card, benchmark and pricing details remain to be verified.
GeneralCoding
0321 Jul 2026
Gemini 3.6 Flash
ImpactMajor
High-signal launch with broad capability, access or platform-level implications.
Google’s newest Flash model, tuned for token-efficient coding, knowledge work and multimodal tasks.
Context1M contextAccessAPIWeightsclosedPriceunknownCheckInferred
Availability
StatusTrackedChannelAPIWeightsClosedPricingUnclearDocsLinked
Source qualityFirst-party model card
Primary model documentation is available and linked from this brief.
Strong source · 1 receiptWhat changed
Capability expansioncontext signalAPI access
Why it matters
Pushes a fast model into long-horizon agents and computer-use workflows without Pro-tier pricing.
GeneralCodingMultimodal
0416 Jul 2026
Grok 4.5
ImpactMajor
High-signal launch with broad capability, access or platform-level implications.
A fast frontier model built around coding, agentic execution and professional knowledge work.
ContextunknownAccessAPIWeightsclosedPrice$2 / $6 per 1MCheckInferred
Availability
StatusTrackedChannelAPIWeightsClosedPricingKnownDocsSource
Source qualityLaunch announcement
Release is backed by the lab’s own announcement or newsroom post.
First-party source · 1 receiptWhat changed
Developer capabilitypricing visibleAPI access
Why it matters
Pairs deep software-engineering ability with an advertised serving speed of 80 tokens per second.
GeneralCoding
0509 Jul 2026
GPT-5.6
ImpactMajor
High-signal launch with broad capability, access or platform-level implications.
A three-tier family: Sol for frontier work, Terra for balance and Luna for fast, efficient tasks.
ContextunknownAccessAPIWeightsclosedPriceunknownCheckInferred
Availability
StatusTrackedChannelAPIWeightsClosedPricingUnclearDocsSource
Source qualityLaunch announcement
Release is backed by the lab’s own announcement or newsroom post.
First-party source · 1 receiptWhat changed
Capability expansionAPI access
Why it matters
Brings stronger multi-agent work and programmatic tool coordination across ChatGPT, Codex and the API.
GeneralCodingMultimodal
0609 Jul 2026
Muse Spark 1.1
ImpactNeeds review
Potentially important, but source strength or availability still needs editorial confirmation.
Meta’s multimodal reasoning model for agentic tasks, computer use, coding and rich media understanding.
Context1M contextAccessAPIWeightsclosedPriceunknownCheckreview
Availability
StatusNeeds checkChannelAPIWeightsClosedPricingUnclearDocsSource
Source qualityNeeds review
Tracked on the radar while at least one comparable field awaits stronger source confirmation.
Editorial review · 1 receiptWhat changed
Capability expansioncontext signalAPI access
Why it matters
Marks Meta’s return to frontier APIs with native multi-agent orchestration and active context management.
GeneralCodingMultimodal
0708 Jul 2026
Robostral Navigate
ImpactNeeds review
Potentially important, but source strength or availability still needs editorial confirmation.
An embodied navigation model that follows language instructions using a single ordinary RGB camera.
ContextunknownAccessresearchWeightsclosedPriceunknownCheckreview
Availability
StatusNeeds checkChannelResearchWeightsClosedPricingUnclearDocsSource
Source qualityNeeds review
Tracked on the radar while at least one comparable field awaits stronger source confirmation.
Editorial review · 1 receiptWhat changed
New interface
Why it matters
Shows a compact model moving beyond screens into autonomous navigation across real environments.
EmbodiedMultimodal
0830 Jun 2026
Claude Sonnet 5
ImpactMajor
High-signal launch with broad capability, access or platform-level implications.
Anthropic’s fifth-generation Sonnet model for autonomous agents, coding, tool use and knowledge work.
ContextunknownAccessAPIWeightsclosedPrice$2 / $10 introCheckInferred
Availability
StatusTrackedChannelAPIWeightsClosedPricingKnownDocsSource
Source qualityLaunch announcement
Release is backed by the lab’s own announcement or newsroom post.
First-party source · 1 receiptWhat changed
Capability expansionpricing visibleAPI access
Why it matters
Brings performance close to Opus 4.8 at a lower price, with stronger planning and long-running execution.
GeneralCodingMultimodal
0916 Jun 2026
GLM-5.2
ImpactMajor
High-signal launch with broad capability, access or platform-level implications.
An open flagship model designed to hold context and execute over long software-engineering horizons.
Context1M contextAccessopenWeightsopenPriceunknownCheckInferred
Availability
StatusTrackedChannelWeightsWeightsOpenPricingUnclearDocsSource
Source qualityLaunch announcement
Release is backed by the lab’s own announcement or newsroom post.
First-party source · 1 receiptWhat changed
Weights openedcontext signalopen weights
Why it matters
Combines a stable million-token window with multiple reasoning-effort levels and openly released weights.
GeneralCodingOpen weights
1009 Jun 2026
North Mini Code
ImpactMajor
High-signal launch with broad capability, access or platform-level implications.
Cohere’s first agentic coding model, built to run locally with a small active compute footprint.
Context256K contextAccessopenWeightsopenPriceunknownCheckInferred
Availability
StatusTrackedChannelWeightsWeightsOpenPricingUnclearDocsSource
Source qualityLaunch announcement
Release is backed by the lab’s own announcement or newsroom post.
First-party source · 1 receiptWhat changed
Weights openedcontext signalopen weights
Why it matters
Makes sovereign coding agents practical: 30B total parameters, but only 3B active at inference time.
CodingOpen weights
1109 Jun 2026
Claude Fable 5
ImpactMajor
High-signal launch with broad capability, access or platform-level implications.
Anthropic’s fifth-generation frontier model for sustained coding, science, vision and demanding knowledge work.
ContextunknownAccessAPIWeightsclosedPrice$10 / $50 per 1MCheckInferred
Availability
StatusTrackedChannelAPIWeightsClosedPricingKnownDocsSource
Source qualityLaunch announcement
Release is backed by the lab’s own announcement or newsroom post.
First-party source · 1 receiptWhat changed
Capability expansionpricing visibleAPI access
Why it matters
Introduces a capability tier above Opus for complex, days-long and asynchronous tasks.
GeneralCodingMultimodal
1201 Jun 2026
MiniMax M3
ImpactMajor
High-signal launch with broad capability, access or platform-level implications.
A natively multimodal frontier model for coding and agents, built on MiniMax Sparse Attention.
Context1M contextAccessopenWeightsopenPriceunknownCheckInferred
Availability
StatusTrackedChannelAPIWeightsOpenPricingUnclearDocsSource
Source qualityLaunch announcement
Release is backed by the lab’s own announcement or newsroom post.
First-party source · 1 receiptWhat changed
Weights openedcontext signalopen weightsAPI access
Why it matters
Combines open weights, a million-token context window and long-horizon agentic execution in one model.
GeneralCodingMultimodalOpen weights
1324 Apr 2026
DeepSeek-V4
ImpactMeaningful
Material model update, access expansion, pricing shift or comparable capability change.
A new API generation offered as V4-Pro and V4-Flash, with both thinking and non-thinking modes.
ContextunknownAccessAPIWeightsclosedPriceunknownCheckInferred
Availability
StatusTrackedChannelAPIWeightsClosedPricingUnclearDocsLinked
Source qualityProduct changelog
Release is backed by first-party product or API change documentation.
Strong source · 1 receiptWhat changed
Developer capabilityAPI access
Why it matters
Keeps compatibility with both OpenAI- and Anthropic-style interfaces, lowering the cost of switching.
GeneralCoding
Reading the radar
Built for signal,
not scorekeeping.
“Latest” means the newest named model announcement we could verify from each lab’s own newsroom or documentation. Release dates reflect the original announcement, not later availability updates.
Claims are deliberately attributed to the releasing company. Benchmark results are not normalized across labs, so this index does not rank models or declare a winner.