Lookonchain APP

App Store

Ornith-1.5 launches a new 122B parameter variant: Built on the Qwen 3.5 architecture, it outperforms Qwen 3.8-Flash-Next in partial programming self-tests.

55 minutes ago

Beating AI Express, a news outlet from Dongcha, reports that DeepReinforce, an AI team founded by renowned NLP scholar Li Jiwei, has added a new 122B MoE model to the Ornith-1.5 series. The Ornith-1.5 series was first released back in August this year with three variants: 397B, 35B-A3B, and 9B; this update fills the gap for a mid-sized model. The new model’s weights are now publicly available, with versions offered for FP8, NVFP4, GGUF, and MLX. Ornith-1.5-122B uses the same architecture as Qwen3.5-122B-A10B, with a total of ~122B parameters and roughly 10B activated parameters per inference. It supports a 256K native context window, image understanding, tool calling, and inference. Official self-tests show the 122B model scored 84.5 on SWE-bench Verified (non-hack), outperforming Qwen3.8-Flash-Next’s 82.8. On CyberGym it scored 80 vs. 78, NL2Repo 51.2 vs. 45.2, and SWE-bench Pro (no-hack) 76.9 vs. 43.9. However, it still lags on Terminal-Bench 2.1 with scores of 81.2 vs. 83.4. Notably, Qwen’s official team previously scored 62.5 on SWE-bench Pro using Claude Code Harness, higher than Ornith’s 43.9 measured via OpenHands, as the two test conditions are not fully aligned.

Relevant content

Fasnara Capital holds short positions worth $192 million, accounting for 86.3% of its total positions, and is currently facing an overall unrealized loss of $5.63 million.

According to monitoring by Onchain Lens, Fasnara Capital’s total position size on Hyperliquid is approximately $222 million, of which 86.3% are short positions valued at around $192 million. The account currently has an overall unrealized loss of about $5.63 million. Its main positions include: BTC short positions worth $51.53 million, with an unrealized loss of $774,500; ETH short positions valued at $51.15 million, with an unrealized gain of $1 million; ZEC short positions worth $26.33 million, with an unrealized loss of $1.39 million; and SKHY long positions worth $21.68 million, with an unrealized loss of $2.64 million. The account has generated a profit of $10.08 million over the past 30 days, with a cumulative historical profit of approximately $122 million.

1 seconds ago

Harmony will shut down its block explorer on November 18.

Harmony announced that it will shut down its block explorer explorer.harmony.one on November 18, 2026. The project team will provide a query tool at migrate.country, which supports looking up address activities and transaction records dating back to January 1, 2025.

1 seconds ago

JPMorgan raises SpaceX's target price from $240 to $500, maintains an overweight rating.

JPMorgan Chase has lifted SpaceX’s price target from $240 to $500, while maintaining an "Overweight" rating. The bank now projects SpaceX’s 2027 revenue will reach $254 billion, its computing power capacity will hit 10 gigawatts (GW) by year-end, and has more than doubled its related forecasts for 2027 to 2028. JPMorgan also views a potential merger between SpaceX and Tesla as strategically significant, though it remains uncertain whether such a deal will materialize.

1 seconds ago

A trader spent $8,790 to buy KIMCHI and liquidated their position for an approximate profit of $255,900.

According to Onchain Lens' monitoring, a trader spent $8,790 to buy 19.2 million KIMCHI tokens, then sold the entire position for roughly $264,600, locking in a profit of $255,900. The trader currently holds no KIMCHI, with the sale proceeds amounting to about 30 times the initial principal. The token's average market cap stood at approximately $455,200 at the time of purchase, and around $13.7 million when sold.

1 seconds ago

PUBG maker Krafton open-sources a 30B model, with its bug-fixing capabilities approaching Meta’s, and it leverages technologies from Kimi and DeepSeek at the base layer.

Beating AI News Brief: South Korean gaming giant KRAFTON has released a preview of Raon-1 mini, a Mixture of Experts (MoE) language model trained from scratch. The model has a total parameter count of around 28.8 billion, with only 4.3 billion activated per inference. It is optimized for Korean dialogue, programming, and agent capabilities, supporting a 260,000-token context window. Though built from scratch, its underlying technology draws on innovations from Chinese tech firms: it uses a hybrid attention architecture, with 18 of its 25 layers employing a linear attention mechanism adapted from Kimi Delta Attention. The team also simplified DeepSeek’s mHC hyperconnection technology to cut training computational costs. KRAFTON’s test results show Raon-1 mini scored 70.8% in the SWE-Verified code repair test, outperforming Qwen3.6-35B-A3B’s 63.2% and nearly matching Meta Muse Glimmer’s 72.0%. However, it lagged in the Terminal-Bench Lite terminal operation test, scoring 47.0%—well below Qwen3.6-35B-A3B’s 63.3% and Muse Glimmer’s 70.7%. For instruction following test IFBench, it hit 84.2%, higher than both comparison models. The model weights are open-sourced under the Apache 2.0 license, with versions available in GGUF, MLX, and other formats, supporting local run on Mac. The NVFP4 quantized version can be deployed on a single RTX 5090. The team also offers the DFlash2 speculative decoding model, which in tests on a single NVIDIA B300 delivers up to 2.18x higher output throughput than standard decoding.

1 seconds ago

Well-known trader: Bitcoin could rise to $86,000 in the short term, with a subsequent target of $94,000.

Well-known trader Doctor Profit posted that he believes Bitcoin is in a bull market, projecting it will surge rapidly to $86,000 in the coming hours to days before climbing further to $94,000. He noted the market is not yet prepared for this move.

1 seconds ago

Popular tokens

BitcoinEthereumHyperliquidSolanaTRONBNBTetherAaveXRPPepeFartcoinOndoJupiterUniswapBonkPendleEthenaArbitrumAvalancheLidoChainlinkPolygonDogecoinCardano