Claude Mythos 5.1
C Tier · 6.5/10
**From 2026-10-06 Mythos 5.1 is included in every tier of Anthropic's expanded Cyber Verification Program, and existing Project Glasswing members move into its Specialized Access tier without reapproval -- with the first disclosed Glasswing numbers: at least 129,000 verified vulnerabilities found April-July 2026, more than 33,000 rated critical or high.** Anthropic's trusted-access frontier model. **Mythos 5.1 launched 2026-09-01** alongside Fable 5.1, and Anthropic now states outright that they are **the same model with different safeguards** -- the 5.1-cycle gap is 60.9% vs 55.8% on Terminal-Bench 4.0, which Anthropic attributes to safeguard interventions rather than capability and expects to shrink. Originally launched June 9, 2026 alongside Claude Fable 5. Suspended June 12 by a US export-control order, then PARTIALLY RESTORED July 1, 2026 (US government lifted controls June 30): Mythos 5 is back for a set of US organizations with government approval, while Anthropic works to re-expand the broader Glasswing program. Public Fable 5 returned globally the same day. Gated to Project Glasswing orgs + select biology researchers.
Score Breakdown
Personality & Tone
The gated red-team specialist
Tone: When Anthropic does publish Mythos outputs (in sanitized research reports), the voice is careful, technically dense, and deliberately unperformed -- much more 'senior security researcher writing an internal memo' than Claude Opus's conversational style.
Quirks: Mythos is tuned to produce its cybersecurity reasoning with extensive show-your-work traces. Anthropic publishes some outputs with full CoT visible as evidence of capability claims. Outside of security tasks, the model reportedly sounds much like Opus 4.6 / 4.7 -- Anthropic hasn't published a distinct general-purpose voice for Mythos.
The Good and the Bad
What we like
- +The most capable Anthropic model available -- meaningfully stronger than Opus 4.7 on cybersecurity reasoning, long-horizon autonomy, and multi-step attack/defense planning per Anthropic's published evaluations
- +73% success rate on expert-level Capture-the-Flag tasks -- a benchmark other frontier models (GPT-5.x, Gemini 3.1 Pro, Opus 4.7) are well below
- +Autonomously executes 32-step network attacks in Anthropic's red-team evals -- demonstrates sustained agentic capability on security tooling without losing track
- +Paired with Project Glasswing: a coalition model where 8 founding enterprise partners get controlled access, $100M in credits, and shared threat intelligence
What could be better
- −Not available to the public. If you're reading this thinking you might use it: you probably can't. Invite-only rollout to ~50 orgs with active cybersecurity or research commitments
- −Even if you are in a Glasswing partner org, access is heavily gated -- deployment requires explicit use-case approval and extensive safety review
- −Specialized for security work. Anthropic explicitly notes Mythos is 'less broadly capable' than Opus 4.7 outside the cyber domain -- so it is NOT the answer for general coding, writing, or analysis work
- −Anthropic withholding the weights and API access is a policy call, not a technical one. This is the first time a frontier Claude model has been deliberately kept out of the API, signaling a new safety/release posture you should expect to see repeat
Pricing
Project Glasswing (Gated)
- ✓Mythos 5 (launched 2026-06-09) restricted to Project Glasswing partners -- expanded to ~150 orgs as of 2026-06-02 -- plus select biology researchers
- ✓Mythos Preview users can upgrade to Mythos 5 immediately
- ✓Founding partners: Amazon, Apple, Google, Cisco, CrowdStrike, JPMorgan, Microsoft, Nvidia
- ✓Broader trusted-access program for cybersecurity + biomedical research planned
- ✓Mandatory 30-day retention on all Mythos-class traffic (not used for training)
Public access (via Claude Fable 5)
- ✓Fable 5 (2026-06-09) is the same underlying model 'made safe for general use' -- classifiers route cyber/bio/chem requests to Opus 4.8 (<5% of sessions)
- ✓Included on Claude Pro/Max/Team/Enterprise at no extra cost through 2026-06-22, usage credits after
- ✓See /tools/claude for the full Fable 5 review
Known Issues
- PROJECT GLASSWING FOLDS INTO THE EXPANDED CYBER VERIFICATION PROGRAM -- MYTHOS 5.1 IN ALL THREE TIERS, GLASSWING MEMBERS BECOME 'SPECIALIZED ACCESS' (2026-10-06, vendor-primary): Anthropic's expanded Cyber Verification Program (CVP) now 'consists of three access tiers' -- Defense Access, Red Team Access and Specialized Access -- and 'each tier includes access to our most capable models, including Claude Opus 5.5, Claude Sonnet 5.5, Claude Mythos 5.1, and new models moving forward'. For this page the material facts are: **(1)** Anthropic is 'integrating these programs into one expanded offering' -- Glasswing (which 'gave a group of organizations securing the most critical software access to Claude Mythos') and the earlier CVP are merged; **(2)** 'Existing members of Project Glasswing will transition to this tier [Specialized Access] and do not require reapproval for current models' -- Specialized Access is reviewed 'in depth in collaboration with the US government' and reserved for organizations testing flight systems, power grids, telecom networks, interbank infrastructure and government networks; **(3)** the Mythos access route for ordinary security teams is now the Defense and Red Team tiers, which Anthropic expects to approve in 'a few days' and 'a few weeks' respectively (Red Team is organizations only, with real-time blocks on physically harmful or mass-disruption actions kept in place); **(4)** data retention is mandatory in CVP, except that organizations already holding zero-data-retention access to Fable 5.1 or Mythos 5.1 can use CVP with ZDR until Enterprise Frontier Safeguards arrives 'later this fall'. **First Glasswing impact figures:** 'at least 129,000 verified software vulnerabilities between April and July 2026' from partners, 5,500 more from Anthropic's own open-source scanning (April to October), 'more than 33,000' rated critical or high, 'likely an undercount' drawn from 33 partner reports; partners said Mythos 'increased their rate of vulnerability finding by months or even years' (Booz Allen and Comcast are cited). Anthropic's Red Team tier test on Opus 5.5 completed 34 of 50 CyScenarioBench tasks with no blocks, matching the no-safeguards rate it treats as representative of Specialized Access. This supersedes the 'US organizations with government approval' framing this page has carried since July 1 as the primary access path; see the Claude page for the full tier definitions.Source: Anthropic (anthropic.com/news/cyber-verification-program, on-page 'Oct 6, 2026', title 'Expanding the Cyber Verification Program') -- fetched 2026-10-06 via curl with browser UA · 2026-10-06
- MYTHOS 5.1 SHIPS -- AND ANTHROPIC NOW SAYS IN WRITING THAT MYTHOS AND FABLE ARE THE SAME MODEL (2026-09-01, vendor-primary): Anthropic launched **Claude Mythos 5.1** alongside the generally-available **Claude Fable 5.1**, and stated the relationship explicitly for the first time: '**Claude Fable 5.1 and Claude Mythos 5.1 are the same model, but with different levels of safeguards. Fable 5.1 is generally available, while Mythos 5.1 is available only through our trusted access programs; its safeguards are specifically designed to support work in cybersecurity and the life sciences.**' **THIS RESOLVES WHAT THIS PAGE HAS HAD TO INFER SINCE JUNE.** The Mythos line is not a separate, more capable model -- it is the same weights with a looser safeguard configuration, and Anthropic has now published the quantitative gap. **THE MEASURED SAFEGUARD TAX, FOR THE FIRST TIME:** on Terminal-Bench 4.0, Mythos 5.1 scores **60.9%** against Fable 5.1's **55.8%**. Anthropic is unusually direct that this five-point spread is **not capability**: 'the gap between them reflects the tasks on which our earlier, less precise cyber safeguards intervened. With the improvements we're making to these safeguards today, we expect the difference between the models to be much smaller.' **So the headline reason to want Mythos access is shrinking by Anthropic's own account** -- the 5.1 cycle narrows rather than widens the Fable/Mythos gap, because the fix was applied to the safeguards rather than to the model. **WHAT ACTUALLY CHANGES ACCESS:** cybersecurity safeguards now produce **60% fewer false positives**, and Anthropic says **Fable 5.1 -- the public model -- can now be used to discover software vulnerabilities, though not to develop exploits for them.** That is capability moving out of the gated tier and into the general one. For biology, Anthropic has established an access program **developed in partnership with the US government** to reach Mythos 5.1's advanced biology capabilities, with enrollment to 'open for scientists soon' -- still gated, but a different and more formal gate than the Project Glasswing arrangement this page has tracked. **PRICING IS IDENTICAL TO FABLE 5.1** and is published on the public rate card: $10/MTok input, $12.50 5-minute cache write, $20 1-hour cache write, **$0.25/MTok cache hits** (a 0.025x multiplier, versus 0.1x on every other Claude model), $50/MTok output; Batch $5/$25. Mythos 5.1 is listed as 'limited availability' via anthropic.com/glasswing. **The practical read: if you were pursuing Mythos access purely for benchmark headroom, the case is weaker than it was in June; if you were pursuing it for cyber or life-sciences scope, the gate is still there but the public model now covers vulnerability discovery.**Source: Anthropic (anthropic.com/claude-fable-and-mythos-5-1 -- post is at the site ROOT, not under /news/) + platform.claude.com/docs/en/about-claude/pricing.md -- both fetched 2026-09-05 via curl with browser UA · 2026-09-01
- PARTIALLY RESTORED (2026-07-01): The US government **lifted the export controls on June 30**. Anthropic redeployed the public **Fable 5 globally on July 1**, but **Mythos 5 came back only partially** -- 'restored access to Mythos 5 for a set of US organizations, following the US government's approval.' Anthropic says it is continuing efforts to 'expand access to the broader set of domestic and international partners in the Glasswing program,' so non-US and many international Glasswing partners may still be waiting. The re-launch shipped a new safety classifier that blocks the reported jailbreak technique in >99% of cases. Net: Mythos 5 is available again, but on a narrower, government-approved US footprint than before the suspension.Source: Anthropic (anthropic.com/news/redeploying-fable-5), CNBC (2026-06-30) · 2026-07-01
- MYTHOS PREVIEW RETIRED (2026-06-30): Per Anthropic's deprecations page, **Claude Mythos Preview reached its retirement date on June 30, 2026** -- Glasswing partners still on the Preview snapshot must now use `claude-mythos-5`. With Fable 5 public and Mythos 5 live, the Preview era is formally closed.Source: Anthropic model deprecations page (platform.claude.com/docs/en/about-claude/model-deprecations) · 2026-06-30
- ACCESS SUSPENDED BY US GOVERNMENT (2026-06-12; RESOLVED 2026-07-01 -- see restoration entry above): A US government export-control directive ordered Anthropic to **suspend all access to Claude Mythos 5 and Claude Fable 5** for any foreign national (inside or outside the US, including foreign-national Anthropic employees). To comply, Anthropic **disabled both models for all customers** -- including, per the directive's scope, Project Glasswing partners. Access to all other Anthropic models is unaffected. The Commerce Department acted after another company claimed it had 'jailbroken' Mythos, raising national-security concerns; Anthropic disagrees and met with the Trump administration on 2026-06-15 to contest the order. This is the first frontier model pulled from access by US-government directive -- a landmark moment for AI export control and the open-vs-closed debate (CNBC framed it as 'a big moment for open-source AI'). Watch for restoration or appeal terms.Source: Anthropic (anthropic.com/news/fable-mythos-access), CNBC (2026-06-12, 2026-06-15, 2026-06-16), TechCrunch, Axios · 2026-06-12
- RETIREMENT DATE SET (verified 2026-06-11): **Claude Mythos Preview retires June 30, 2026** per Anthropic's deprecations page -- Glasswing partners still on the Preview snapshot must migrate to claude-mythos-5 before then. With Fable 5 public and Mythos 5 live, the Preview era formally closes out at the end of the monthSource: Anthropic model deprecations page (platform.claude.com/docs/en/about-claude/model-deprecations) · 2026-06-11
- MODEL LAUNCH (2026-06-09): **Claude Mythos 5** replaces Mythos Preview as the Glasswing-track model -- same underlying model as the publicly available Claude Fable 5, with safeguards lifted in certain areas for vetted partners. Existing Mythos Preview users upgrade immediately. Anthropic simultaneously made the Mythos class public for the first time via Fable 5 ($10/$50 per 1M, plan-included through 6/22), with classifier fallback to Opus 4.8 on cybersecurity, bio/chem, and distillation-attempt requests. All Mythos-class traffic now carries mandatory 30-day retention, overriding zero-data-retention agreements (not used for training)Source: Anthropic news (anthropic.com/news/claude-fable-5-mythos-5), TechCrunch, CNBC · 2026-06-09
- Mythos's cybersecurity capability is the reason for its gated release. Anthropic's red-team evaluations showed the model could plan end-to-end network intrusion chains, which Anthropic deemed too risky for open API accessSource: Anthropic Project Glasswing announcement, Axios, CNBC, Schneier on Security · 2026-04
- Naming history: 'Claude Mythos Preview' was the April 2026 product name (internal codename Capybara); the 'Mythos 5' name became official with the 2026-06-09 launch, aligning with Fable 5 (there was never a Mythos 1-4)Source: Axios, Fortune, Anthropic news · 2026-06
- Access applications are not open -- Anthropic is approaching partner orgs directly rather than accepting inbound requestsSource: Anthropic Glasswing page · 2026-04
- Axios reported 2026-04-19 that the NSA is among the ~40 orgs with Mythos access, despite the Pentagon's formal supply-chain risk designation of Anthropic. Dario Amodei reportedly met with W.H. Chief of Staff Susie Wiles and Treasury Secretary Scott Bessent on 2026-04-17. Material context if you are evaluating Mythos / Glasswing in a federal or defense-adjacent procurement -- the political posture inside the US government is not uniformSource: Axios, TechCrunch, Engadget · 2026-04
Best for
Partner organizations in Project Glasswing doing cybersecurity research, defensive red-teaming, threat intelligence, or large-scale vulnerability triage. If your use case is legitimate cybersecurity and you have enterprise Anthropic contact, ask about Glasswing admission.
Not for
Everyone else -- but as of June 9, 2026 'everyone else' gets Claude Fable 5 (see /tools/claude): the same Mythos-class model made safe for general use, on the API and included in paid plans through June 22.
Our Verdict
UPDATE (July 1, 2026): the suspension is over. The US government lifted the export controls on June 30; public Fable 5 returned globally on July 1, and Mythos 5 was restored -- but only partially, to a set of US organizations with government approval, while Anthropic works to re-admit the broader (and international) Glasswing base. So Mythos 5 is available again, on a narrower footprint than before, and Mythos Preview formally retired June 30. The pre-suspension picture, for context: the Mythos story changed on June 9, 2026. What began in April as a deliberately withheld cybersecurity preview is now a two-track release: Mythos 5 for ~150 vetted Glasswing orgs and select biology researchers with safeguards lifted, and Claude Fable 5 for the public -- the same model with classifier-enforced fallbacks to Opus 4.8 on dangerous-capability requests. That makes this page's subject the gated track only. If you're in Glasswing, Mythos 5 is an immediate upgrade from Mythos Preview. If you're not, you no longer have to wonder what you're missing: Fable 5 IS the Mythos class, minus the <5% of sessions that touch cyber/bio/chem territory. The deeper signal stands -- Anthropic now ships its frontier in safety-differentiated tiers, and the 30-day mandatory retention on all Mythos-class traffic shows what public access to this capability level costs in privacy terms.
Sources
- Anthropic: Expanding the Cyber Verification Program -- Glasswing members transition to Specialized Access, Mythos 5.1 in every tier, 129,000 vulnerabilities (2026-10-06) (accessed 2026-10-06)
- Anthropic: Introducing Claude Fable 5.1 and Claude Mythos 5.1 (2026-09-01) -- same model, different safeguards (accessed 2026-09-05)
- Anthropic pricing docs: Mythos 5.1 at $10/$50 with $0.25/MTok cache hits, limited availability via Glasswing (verified 2026-09-05) (accessed 2026-09-05)
- Anthropic: Redeploying Fable 5 (Mythos 5 partially restored 2026-07-01) (accessed 2026-07-04)
- Anthropic: Statement on the US government directive to suspend access to Fable 5 and Mythos 5 (2026-06-12) (accessed 2026-06-18)
- CNBC: Anthropic's Fable shutdown is a big moment for open-source AI (2026-06-16) (accessed 2026-06-18)
- Anthropic: Introducing Claude Fable 5 and Claude Mythos 5 (2026-06-09) (accessed 2026-06-09)
- Anthropic: Project Glasswing (accessed 2026-04-17)
- Anthropic Red: Mythos Preview (accessed 2026-04-17)
- Fortune: Anthropic's Mythos model + Project Glasswing (accessed 2026-04-17)
- Axios: Anthropic releases Opus 4.7, concedes it trails unreleased Mythos (accessed 2026-04-17)
- Schneier on Security: On Mythos Preview and Project Glasswing (accessed 2026-04-17)
- CNBC: Anthropic Opus 4.7 less risky than Mythos (accessed 2026-04-17)
- Axios: NSA uses Mythos despite Pentagon feud (2026-04-19) (accessed 2026-04-20)
- TechCrunch: NSA spies reportedly using Anthropic Mythos (accessed 2026-04-20)
Explore more Claude Mythos 5.1 rankings
Deeper leaderboards, benchmarks, task-specific tier lists, and status/pricing pages for Claude Mythos 5.1.
The Tier List Tuesday
Weekly newsletter: tier movers, new entrants, and the VS of the week. Built from our daily AI-tool sweeps. No spam, unsubscribe anytime.
Alternatives to Claude Mythos 5.1
Claude (Anthropic)
**Claude Haiku 5.5 shipped on 2026-10-07 at $0.10 / $0.50 per 1M tokens for prompts up to 100K ($0.50 / $2.50 above) -- 90% below Haiku 4.5 on most requests -- with adjustable effort for the first time on a Haiku, and the same day Anthropic halved Sonnet 5.5 cache reads to $0.10 and added monthly API credits for Max ($100 or $200) and Team (up to $500 pooled).** **Anthropic expanded the Cyber Verification Program on 2026-10-06 into three access tiers (Defense, Red Team, Specialized) that unlock reduced cyber safeguards on Opus 5.5, Sonnet 5.5, Mythos 5.1 and future models for verified security teams -- closing the 'coming weeks' note Opus 5.5 shipped with; generally available Claude keeps its conservative cyber blocks.** **Two Claude 5.5 models in six days: Claude Opus 5.5 (2026-09-22) and Claude Sonnet 5.5 (2026-09-28).** Opus 5.5 performs at Fable 5.1 level on most work and costs 40% less than Opus 5 to run -- $4/$20 per 1M (down from $5/$25) with cache reads cut to $0.20 (from $0.50), fast mode at $8/$40 up to 2.5x speed, output 30%+ faster, and five-hour usage limits raised on Pro, Max, Team and seat-based Enterprise with a saveable rate-limit reset. Sonnet 5.5 keeps Sonnet 5's $2/$10 but needs far fewer tokens (up to 30% cheaper per task), runs 30%+ faster and jumps Terminal-Bench 4.0 from 10.3% to 70.6%. Both are 1M context, Jun 2026 cutoff, on the Claude Platform, Bedrock, Vertex and Foundry; Haiku 5.5 followed on 2026-10-07. Opus 5.5 launches with Fable-class cyber and bio safeguards (most cyber tasks fall back to Opus 4.8) and cannot run with thinking off.
Gemini (Google)
**Gemini 4 Argon went generally available on 2026-10-13: in the Gemini API (paid tier only), AI Studio, Gemini for business and GitHub Copilot, in the Gemini app and AI Mode in Search for Google AI Pro and Ultra subscribers, and in Google Antigravity for Ultra. The API rate card prices it at $2 input / $10 output per 1M tokens with $0.10 implicit-cache reads through November 30, 2026, doubling to $4 / $20 / $0.20 on December 1; there is no free tier and no explicit caching.** Google calls it its flagship reasoning model, with a 1M-token unified input-output limit, #1 on the Vals Index, 87% on the ERQA embodied-reasoning benchmark and 0.7% attack success on Gray Swan prompt injection, alongside the 9/30 numbers (77.9% DeepSWE v1.1, 51.3% AutomationBench, 91.7% LVBench, 68% CWE-bench v1) -- all vendor-reported. Earlier in October: Gemini 3.7 Flash and 3.5 Flash were retired from AI Studio (10/08) with API calls auto-routed to 3.8 Flash and 3.6 Flash, Nano Banana 2.1 went GA (10/06) and SynthID Detector opened to everyone in English (10/07). Late September in the Gemini app: **skills replace Gems** (9/30; slash-command instructions with reference files, Gems removed from November 2026 for personal accounts), **Guided Vision in Gemini Live** for blind and low-vision users on Android (10/01), and the 9/23-9/24 wave still stands: Gemini 3.8 Flash TTS and Flash-Lite TTS with voice design at $0.50/$9.00 and $0.50/$6.00 per 1M (doubling 2027-01-01), Gemini 3.8 Live with Live Avatar in Gemini Enterprise, free 1080p Omni 1.1 video in Google Vids. Googlebook was scheduled to reach shelves 10/04 (shipment not yet vendor-confirmed).
Grok
**The X Subscription (2026-10-12) folds Grok, Grok Bot, Cursor and X Premium into one plan: Free, Plus $8, Premium $30 (what SuperGrok, Cursor Pro and X Premium+ cost $90 for combined), Super $100 and Ultra $200 a month, with one weekly-refreshing usage pool shared across Grok and Cursor; the old SuperGrok, Cursor and X Premium plans are closed to new sign-ups.** **Team Bots launched 2026-09-28 in public beta on Teams and Enterprise plans** -- shared Grok Bots built around a role, with team context (files, instructions, skills), plugins (Salesforce, Notion, GitHub), credentials for APIs without a plugin, and memories, where the bot is shared but each person's conversations and memories stay private; each Team Bot gets its own Slack handle. SpaceXAI runs one per major sales account, and its 9/22 case study says the merged SpaceXAI + Cursor support desk absorbed a 175% ticket increase on Grok Bot with no new hires at $0.20-$0.30 per resolution. **Grok 4.7 (9/21)** remains the flagship at $2/$6 per 1M with chart-only benchmarks; Grok Voice Transcribe 2.0 and the corrected $15/1M-character TTS price stand.
Muse Spark + Muse Wonder (Meta)
Meta Superintelligence Labs' model family, now four lines deep: **Muse Wonder** (2026-10-12, frontier, $4/$20 per 1M on the Meta Model API; Wonder Full only in Muse Code subscriptions), **Muse Spark 1.3** (2026-09-02, $1.25/$4.25, or $0.10/$0.20 in Contributor mode if Meta may train on your content), **Muse Spark Open** (2026-10-15, 816B/42B-active open weights, 1M context, runs on two Mac Studios with M5 Ultra) and **Muse Glimmer** (2026-08-10, 30B, Apache 2.0, built for local agents). Consumers get it through **Muse**, Meta's personal-agent app (2026-09-08): free with limits, Power $20/mo, Maximum $100/mo
GPT-Rosalind (OpenAI)
**Since 2026-10-15 US university labs can apply for the Rosalind Academic Plan: up to 30 ChatGPT Business Standard seats free for up to 12 months, with GPT-Rosalind and Rosalind Workbench behind trusted-access approval.** OpenAI's first domain-specific model -- life sciences, drug discovery, translational medicine. Launched 2026-04-16 as a Trusted Access research preview. Launch partners: Amgen, Moderna, Allen Institute, Thermo Fisher. Paired with a Life Sciences Codex plugin (50+ scientific tool integrations)
GPT-5.6-Cyber / GPT-5.4-Cyber (OpenAI)
**GPT-5.4-Cyber reached its API removal date on 2026-10-01: the gpt-5.4-cyber model page now returns Not found while gpt-5.6-cyber stays live (checked 2026-10-02; OpenAI's deprecations doc had not yet moved the row to Past Deprecations). Its replacement is `gpt-5.6-cyber`**, an alias for OpenAI's most advanced purpose-trained cyber models, gated behind the Daybreak program and priced at $12.50/$75 per 1M (first published price for any OpenAI cyber model). Original page: OpenAI's defensive-cybersecurity variant of GPT-5.4, launched 2026-04-16. Lowered refusal boundary for security-research tasks and native binary reverse-engineering. Access gated via Trusted Access for Cyber (TAC) program -- thousands of verified defenders, hundreds of teams, no public pricing. On **2026-08-17 OpenAI published its first dedicated post on the Hugging Face incident**, conceding it 'underestimated the real-world cyber capabilities of our AI models' and confirming it now releases cyber capabilities only to trusted defenders. **On 2026-09-01 OpenAI confirmed GPT-6 Astra meets the Critical cyber threshold -- the first model it has ever designated at that level -- and on 2026-09-03 committed $1B to Daybreak for Frontline Defenders**
Microsoft MAI-Thinking-1
Microsoft's first in-house reasoning model -- launched 2026-06-02 at Build as the flagship of seven new MAI models. 35B-active / ~1T-total sparse Mixture-of-Experts, 256K context. AIME 2025 97.0%, matches leading models on SWE-Bench Pro, and beat Claude Sonnet 4.6 in human-preference testing. Available on Microsoft Foundry + OpenRouter / Fireworks / Baseten
Hunyuan 3 (Tencent Hy3)
Tencent's Hy3 reached GA 2026-07-06 (upgraded from the April preview) -- 295B total / 21B active MoE, 256K context, now Apache 2.0 open weights on HuggingFace + ModelScope with the EU/UK/South Korea restriction lifted. ~90% agent-task completion on Tencent's internal apps; API via Tencent Cloud TokenHub. Integrated into Yuanbao, WeChat, QQ
MiMo (Xiaomi)
Xiaomi's MiMo-V2.5 family launched 2026-04-22 -- Pro (1T total / 42B active MoE, 1M context, native vision+audio reasoning), Multimodal base, TTS (3 sub-models: base, VoiceDesign, VoiceClone), and ASR (open-source, English + Chinese + major dialects). Full voice pipeline for the agent era. Extra-charge 1M-context tier removed at launch