Qwen 3.8-27B Surpasses Muse Glimmer, Approaches Claude Opus 4.6
By: t.me|2026/08/14 15:25:58
0
Share
Prefer us on Google
Qwen has officially open-sourced Qwen 3.8-27B, with official benchmark scores showing that this model outperforms Meta's 30B model Muse Glimmer in all 8 directly comparable tests. Qwen 3.8-27B scored 73.0 in Terminal-Bench 2.1, while Muse Glimmer scored 51.7; in SWE-bench Pro, it scored 61.7 compared to 51.2; and in OSWorld-Verified, it achieved 84.3 against 65.9. It also leads in general reasoning, document, and visual tests. In comparison with Claude Opus 4.6, Qwen 3.8-27B won 15 out of 19 tests, surpassing in SWE-bench Pro, LiveCodeBench, OSWorld, AndroidWorld, and several visual tasks, while still trailing in Terminal-Bench, GPQA, HLE, and NL2Repo. The official benchmark scores of this 27B model are now close to the level of the previous generation's closed-source flagship.
-- Price
--
--
--
This content is provided for general informational purposes only and doesn't constitute financial, investment, legal, or tax advice. Any events, rewards, online promotions, or related information mentioned herein should not be considered a recommendation, solicitation, or invitation to purchase, sell, trade, or otherwise deal in any crypto assets. Crypto assets are highly volatile and may result in loss. The availability of WEEX services, products, and related events may vary by region. You are responsible for ensuring that your participation is in accordance with applicable local laws and regulations.
You may also like
Fake Claude App 101MB Attempts to Steal Cryptocurrency Wallets
A malicious campaign to extract cryptocurrency wallets and account information through fake desktop installation files has been confirmed at the time of Anthropic's release of the new Claude model.
Fake Claude App Spreads RevStealer, Targeting 50 Crypto Wallets
Fable 5 Expenditure at 11.4%, Changes in AI Adoption
Anthropic's top AI model, Claude Fable 5, accounted for 11.4% of corporate spending in July. Companies are increasingly favoring lower-cost AI models.
Anthropic Reveals Two Claude Models, Marshmallow Offers Better Conversational Experience than Opus 5
Corporate Clients Opt for Cheaper Models Instead of Claude Fable 5
GPT-5.6 Sol Costs 13% Less Than Kimi K3
The belief that Chinese AI models are cheaper than American ones has been challenged in actual cost comparisons. While the token price is lower for Chinese models, the number of tokens used for answer generation and quality also play a crucial role.
Claude Sonnet 5 to Transition to Standard Pricing on September 1
The price of Anthropic's Claude Sonnet 5 API is set to transition to standard pricing on September 1.
Cases of AI Agent Accessing External Infrastructure
Cases of AI agents accessing actual external infrastructure during internal cybersecurity assessments have been repeatedly confirmed.
Alibaba internally announced a complete ban on Claude, which will take effect on July 10
According to reports from ZhDongxi, Alibaba announced today an全面 ban on the Claude series products under Anthropic. All employees are required to uninstall Agent products, including multiple series models such as Sonnet, Opus, Fable, and Claude Code. This ban will officially take effect on July 10. ...
Zcash vulnerability researcher will audit Monero and plans to expand to more privacy coin projects
According to CoinDesk, security engineer Taylor Hornby, who discovered a serious vulnerability in Zcash using the Anthropic Opus 4.8 AI model, stated that Monero (XMR) has been added to the audit queue, and more privacy coin projects will undergo security reviews in the future.Hornby previously disc...
B.AI subsidy increase: Limited time opening of 1:1 equivalent recharge rebate mechanism
According to official news, B.AI launched a significant recharge benefits upgrade event on May 3, temporarily opening a 1:1 equivalent recharge rebate mechanism. Users can enjoy an equivalent amount of credits upon recharge, achieving an ultra-low cost of 50% for accessing cutting-edge models like G...
Microsoft enables GPT to collaborate with Claude, outperforming all competitors
Microsoft announced on Monday the launch of two new features for Copilot Researcher—Critique and Council—combining OpenAI's GPT with Anthropic's Claude for the same research task.Critique adopts a sequential collaboration model: GPT is responsible for planning the research, retrieving information, a...
OpenClaw has designated Venice AI as the recommended model provider, and the VVV token has risen over 70% in the past 7 days
Founder of Venice AI, Erik Voorhees, posted on the X platform that the OpenClaw documentation shows Venice has been set as the recommended model provider, with the default being venice/llama-3.3-70b and "best overall" being venice/claude-opus-45.Market data shows that Venice Token (VVV) has risen by...
DefiLlama and Forgd introduce institutional token grades, but warn that AAA does not mean risk-free
21-bank stablecoin has global backing, but can it rival USDT and USDC?
Bitcoin faces a two-week Fed trap as inflation rewrite threatens to upend rate cuts
Bitcoin faces a Sept. 16 Fed decision before August PCE and annual revisions land Sept. 30, resetting inflation and rate expectations.
The source on Naura and 3D DRAM memory is unavailable
The provided link to report on an alleged advancement by Naura and CXMT in 3D DRAM memory returns a 404 error. Although there are separate references to CXMT, Naura, and advanced memory developments, none of the accessible sources confirm that Naura has presented its own 3D DRAM technology alongside...
Ubuntu 26.10 Releases Snapshot 3 for Testing Ahead of October Launch
Ubuntu has released Snapshot 3 of the upcoming version 26.10, a monthly ISO intended exclusively for testing as the official launch approaches in October.
Kalshi faces $500,000 daily fines as Michigan forces sports event contracts offline
The Kalshi Michigan injunction imposes a $500,000 daily geofencing penalty and FCM notice duty as federal appeals courts split on preemption.
30-Year Bond Yield Hits 4.079%, Raising Funding Costs for MetaPlanet's Bitcoin Purchases
The average bid yield for Japan's 30-year government bonds has reached 4.079%, raising the cost of yen financing for companies looking to purchase Bitcoin (BTC). Existing fixed-rate debts will not change immediately, but the threshold for borrowing money in the same manner has increased.
Quantum Memory: The Device That Breaks Bitcoin and Replaces It
Japan’s 4% bond yield spike threatens the low-cost borrowing strategy behind corporate Bitcoin buying
Metaplanet BitBonds keep fixed terms, but Japan’s 4% long-bond yield raises the funding hurdle for future Bitcoin purchases and refinancing.
Anthropic's Mea Culpa: A Complete Autopsy of Claude's Missteps
Anthropic details, a month after the incidents, the security and alignment flaws behind Claude's accidental breaches, and strengthens its safeguards.
Arthur Hayes calls EUR/JPY prices crypto’s smoke alarm, but the Fed’s plumbing still shows no fire
EUR/JPY fell 2.38%, but zero FIMA repo use and no scheduled reserve-management purchases show Arthur Hayes’ crypto liquidity thesis has not fired.
Debate Over $300 Bitcoin Tax Exemption and Estimated Revenue Increase
A proposal in the U.S. suggests that allowing a $300 tax exemption for small Bitcoin transactions could increase tax revenue, reigniting the debate over digital asset taxation.
Copy Trading: How Does It Work in 2026?
PL Deputy Proposes Gun Carrying Rights for Cryptocurrency Investors and Industry Executives
Deputy Captain Alden (PL/BA) presents a bill proposing gun carrying rights for investors and executives in the cryptocurrency sector.
The Executive Who Anticipates a New Era for Cryptocurrencies: "We Are Just Getting Started"
Why GENIUS could leave digital dollars vulnerable to sudden blockchain network ‘bank runs’
A Fed paper shows how blockchain congestion can trigger stablecoin redemptions or chain migration despite safe reserves under the GENIUS Act.
Robinhood Chain Down for 14 Minutes: The Blockchain That Was Supposed to Tokenize Wall Street First Blocked Itself
Robinhood Chain stopped producing blocks for over 14 minutes on Friday. Timeline, unknown cause, and what it says about the network.
Fake Claude App 101MB Attempts to Steal Cryptocurrency Wallets
A malicious campaign to extract cryptocurrency wallets and account information through fake desktop installation files has been confirmed at the time of Anthropic's release of the new Claude model.
Fake Claude App Spreads RevStealer, Targeting 50 Crypto Wallets
Fable 5 Expenditure at 11.4%, Changes in AI Adoption
Anthropic's top AI model, Claude Fable 5, accounted for 11.4% of corporate spending in July. Companies are increasingly favoring lower-cost AI models.
Anthropic Reveals Two Claude Models, Marshmallow Offers Better Conversational Experience than Opus 5
Corporate Clients Opt for Cheaper Models Instead of Claude Fable 5
GPT-5.6 Sol Costs 13% Less Than Kimi K3
The belief that Chinese AI models are cheaper than American ones has been challenged in actual cost comparisons. While the token price is lower for Chinese models, the number of tokens used for answer generation and quality also play a crucial role.
...
Customer Support:@weikecs
Business Cooperation:@weikecs
Quant Trading & MM:bd@weex.com
VIP Program:support@weex.com
