AI Model Daily - 2026-06-01 | QName AI

30-Second Summary

Today's most important new signal came from video models: Arena announced inside the window that Grok-Imagine-Video-1.5-Preview (720p) reached #1 on the Image-to-Video Arena, with the public leaderboard showing 1473+/-9, ahead of dreamina-seedance-2.0-720p and happyhorse-1.0.
xAI docs now include a grok-imagine-video-1.5-preview API entry, with alias grok-imagine-video-1.5-2026-05-30, regions us-east-1 / eu-west-1, and 720p output pricing shown as $0.14/s; an xAI employee said it is currently available through the API and will come to app/web soon.
Artificial Analysis posted a leaderboard signal for HiDream-O1-Image-Dev-2604 inside the window: it is now the leading open-weights model on the Artificial Analysis Text to Image Arena, with the public page showing open-weights Elo 1191.
No new general flagship LLM release was visible from OpenAI, Anthropic/Claude, Google DeepMind/Gemini, Meta, Mistral, Qwen, Cohere, Runway, Luma, Kling, or other major official accounts in the last 24 hours; the LLM leaderboard still centers on Claude Opus 4.8 / GPT-5.5.
@xiaohu reposted/commented on a "GPT-Realtime 2.0 real-time voice control of a computer" demo inside the window, but no OpenAI official release or docs were found; this should be treated as a high-heat X lead / unverified rumor, not an official model first release.

Must Watch

1. xAI: Grok Imagine Video 1.5 Preview Reached #1 on Arena Image-to-Video

Event type: Arena appearance + official docs availability + X first signal / employee signal.
First/official time: Arena posted at 2026-05-31 08:00 HKT; the xAI docs page shows model alias grok-imagine-video-1.5-2026-05-30; xAI employee Jie Huang posted at 2026-06-01 02:10 HKT that Grok Imagine 1.5 Preview is available through the API.
Capability change: Arena said Grok-Imagine-Video-1.5-Preview (720p) improved by +52 pt over Grok-Imagine-Video (720p) and surpassed Seedance 2.0 and HappyHorse. The public Image-to-Video Arena page shows grok-imagine-video-1.5-preview-720p at 1473+/-9 in #1, dreamina-seedance-2.0-720p at 1467+/-11 in #2, happyhorse-1.0 at 1443+/-12 in #3, and the original grok-imagine-video-720p at 1421+/-6 in #4.
Availability: xAI API docs list grok-imagine-video-1.5-preview; the model supports Text / Image input to Video output, is available in us-east-1 and eu-west-1, shows base output pricing of $0.08/s, resolution pricing of 480p $0.08/s and 720p $0.14/s, and RPM 60. Jie Huang replied that it is API only for now and will roll out to app soon.
Impact: this is today's most material new model / new entry point. It is not just a leaderboard rumor because Arena leaderboard, Arena X, and xAI docs all corroborate it. The score is still marked Preliminary, and the main @xai account has not posted a standalone announcement, so real quality, stability, moderation behavior, and failure rate still need testing.
Sources: Arena Image-to-Video Leaderboard, Arena X, xAI Grok Imagine Video 1.5 Preview docs, Jie Huang X

2. Artificial Analysis: HiDream-O1-Image-Dev-2604 Is the Leading Open-Weights Text-to-Image Model

Event type: third-party leaderboard update + open model card / repository verification.
First/official time: @ArtificialAnlys posted at 2026-06-01 04:52 HKT; the HiDream Hugging Face model card says HiDream-O1-Image-Dev-2604 was open-sourced on 2026-05-14.
Capability change: Artificial Analysis said HiDream-O1-Image-Dev-2604 is the current leading open-weights Text to Image model in the Artificial Analysis Image Arena. The public FAQ shows the open-weights Text to Image ranking led by HiDream-O1-Image-Dev-2604 with Elo 1191, followed by Qwen Image Max 2512 at 1161 and FLUX.2 [dev] at 1158. The overall Text to Image ranking is still led by GPT Image 2 (high) at 1338.
Availability: Hugging Face HiDream-ai/HiDream-O1-Image-Dev-2604; the model card lists MIT License, 9B params, Image-Text-to-Image / Transformers / Safetensors, and provides AutoProcessor / AutoModelForImageTextToText loading examples. The page says it is not currently deployed by an Inference Provider.
Impact: this matters for open image generation stacks. It was not newly open-sourced today, but it received a strong third-party leaderboard signal today. It should enter local/private image-generation evaluation pools, especially for 2048x2048 generation, long-text rendering, layout control, and prompt-refiner workflows; deployment requirements, VRAM, speed, and real commercial-license boundaries still need separate evaluation.
Sources: Artificial Analysis X, Artificial Analysis Text to Image, HiDream-O1-Image-Dev-2604 on Hugging Face, HiDream-O1-Image model card

3. XiaoHu: GPT-Realtime 2.0 Voice Computer Control Demo Is Worth Tracking, but Not Official

Event type: X-first propagation + unverified rumor.
X time: @xiaohu commented on / reposted Farza's voice-controlled computer demo at 2026-05-31 09:24 HKT.
Capability change: XiaoHu described it as "GPT-Realtime 2.0 real-time voice connecting to AI to control your computer, this is the real Siri"; the original post demonstrated controlling a computer using voice only.
Availability: only the X demo and repost/commentary were visible; this pass found no OpenAI official blog, developer docs, model card, or API changelog confirming the "GPT-Realtime 2.0" name.
Impact: this is highly aligned with OpenAI Realtime / computer use / Codex product direction, but the evidence chain is not enough. Continue watching OpenAI Developers changelog, Realtime API docs, Codex changelog, and the original author's follow-ups; do not treat it as a model release yet.
Sources: XiaoHu X, Farza X

4. General LLMs Today: Major Official Accounts Were Quiet, Leaderboards Extend the Same Pattern

Event type: official-source verification + leaderboard watch.
First/official time: no new OpenAI / Anthropic / Google DeepMind / Gemini / Meta / Mistral / Qwen / Cohere general flagship model release was visible in the last 24 hours.
Capability change: Artificial Analysis LLM still shows Claude Opus 4.8 (max) leading with Intelligence Index 61, followed by GPT-5.5 (xhigh) at 60, GPT-5.5 (high) at 59, and Claude Opus 4.7 (max), Gemini 3.1 Pro Preview, and Qwen3.7 Max in the 57 tier.
Availability: no new general LLM API / web / cloud platform entry point appeared.
Impact: today should not be read as a general LLM launch day. The real increment is video-model API availability and open image-model leaderboard movement.
Sources: Artificial Analysis LLM, LMArena Text

Arena Watch

Arena Image-to-Video: the page is dated 2026-05-29, with 1,155,731 votes and 40 models. The top entries are grok-imagine-video-1.5-preview-720p 1473+/-9 Preliminary in #1; Bytedance dreamina-seedance-2.0-720p 1467+/-11 in #2; Alibaba-ATH happyhorse-1.0 1443+/-12 in #3; xAI grok-imagine-video-720p 1421+/-6 in #4; Google veo-3.1-audio 1397+/-11 in #5 and veo-3.1-audio-1080p 1393+/-10 in #6.
Arena Text-to-Video: the page is still dated 2026-05-29, with 452,799 votes and 17 labs. Labs view still shows Bytedance dreamina-seedance-2.0-720p 1463+/-9 in #1, Alibaba-ATH happyhorse-1.0 1440+/-8 in #2, and Alibaba wan2.7-t2v 1385+/-21 in #3.
LMArena Text Overall: the page remains dated 2026-05-28, with 6,532,755 votes and 360 models. The top entries are still claude-opus-4-6-thinking 1502+/-4, claude-opus-4-7-thinking 1500+/-5, claude-opus-4-6 1498+/-4, claude-opus-4-7 1494+/-5, muse-spark 1489+/-6, gemini-3.1-pro-preview 1487+/-4, gemini-3-pro 1486+/-4, and gpt-5.5-high 1482+/-6. Opus 4.8 still does not have an official visible score there.
Arena Code / WebDev: the page remains dated 2026-05-24, with 328,594 votes and 81 models. The top entries are claude-opus-4-7-thinking 1567+10/-10, claude-opus-4-7 1562+10/-10, claude-opus-4-6-thinking 1542+8/-8, qwen3.7-max-20260517 1541+16/-16, and claude-opus-4-6 1538+7/-7. OpenAI gpt-5.5-xhigh (codex-harness) sits at 1505+10/-10.
Artificial Analysis LLM: Claude Opus 4.8 (max) has Index 61, cost $4.10/M, speed 56 tokens/s; GPT-5.5 (xhigh) has 60, $4.35/M, 61 tokens/s; GPT-5.5 (high) has 59; Claude Opus 4.7 (max) has 57; Gemini 3.1 Pro Preview has 57; Qwen3.7 Max has 57.
GDPval-AA: Claude Opus 4.8 (Adaptive Reasoning, Max Effort) remains first with Elo 1890 and CI -34/+35; GPT-5.5 (xhigh) is 1769, GPT-5.5 (high) 1755, Claude Opus 4.7 (max) 1753, Gemini 3.5 Flash (medium) 1658, Gemini 3.5 Flash (high) 1656, and Qwen3.7 Max 1548.
ITBench-AA: Opus 4.8 still does not appear; the page summary shows Claude Opus 4.7 (max) first at 46.7%, GPT-5.5 (xhigh) second at 45.8%, and Qwen3.7 Max third at 42.5%.
Artificial Analysis Text to Image: the overall ranking is still led by GPT Image 2 (high) Elo 1338, followed by GPT Image 1.5 (high) 1268, Nano Banana 2 (Gemini 3.1 Flash Image Preview) 1260, Nano Banana Pro (Gemini 3 Pro Image) 1219, and Recraft V4.1 Utility Pro 1204. The open-weights subranking is led by HiDream-O1-Image-Dev-2604 at 1191.
Artificial Analysis Image Editing: the overall ranking is still led by GPT Image 1.5 (high) 1260, followed by GPT Image 2 (high) 1252, Nano Banana Pro 1241, Nano Banana 2 Preview 1240, and grok-imagine-image-quality 1231. The open-weights subranking is led by HunyuanImage 3.0 Instruct (Fal) at 1225, with HiDream-O1-Image second at 1206.
Artificial Analysis Image to Video: with audio, Dreamina Seedance 2.0 720p remains first with Elo 1192, HappyHorse-1.0 second at 1169, Veo 3.1 1091, grok-imagine-video 1091, and Veo 3.1 Fast 1084; without audio, HappyHorse-1.0 remains first with Elo 1400, followed by Dreamina Seedance 2.0 720p 1345, grok-imagine-video 1329, PixVerse V6 1326, and Kling 2.5 Turbo 1080p 1288. Artificial Analysis has not yet separated Grok Imagine Video 1.5 Preview here.
Artificial Analysis Text to Video: with audio, Dreamina Seedance 2.0 720p remains first with Elo 1216, HappyHorse-1.0 second at 1213, Kling 3.0 1080p (Pro) 1104, Kling 3.0 Omni 1080p (Pro) 1100, and Veo 3.1 1095; without audio, HappyHorse-1.0 remains first with Elo 1358, followed by Dreamina Seedance 2.0 720p 1272, Kling 3.0 1080p (Pro) 1250, Kling 3.0 Omni 1080p (Pro) 1235, and grok-imagine-video 1234.
Sources: Arena Image-to-Video, Arena Text-to-Video, LMArena Text, Arena Code, Artificial Analysis LLM, GDPval-AA, ITBench-AA, Text to Image, Image Editing, Image to Video, Text to Video

X Watch

@arena: posted inside the window that Grok-Imagine-Video-1.5-Preview (720p) reached #1 on the Image-to-Video Arena, saying it improved by +52 pt over the original Grok-Imagine-Video (720p); follow-up posts linked the Arena leaderboard and xAI docs.
@ArtificialAnlys: posted HiDream-O1-Image-Dev-2604 inside the window, saying it is the leading open-weights Text to Image model; the same thread says HiDream-O1-Image is the second-highest open-weights model on Image Editing Arena.
@xiaohu: had three AI-related signals inside the window: a GPT-Realtime 2.0 voice-computer-control demo that needs official verification; a community comment on users switching between Claude Code and Codex for vibe coding; and a repost of a Chinese image-generation GitHub Skill trending signal, which is a tool/workflow lead rather than a model release.
@xai: no standalone new post was visible on the main account inside the window; the reliable Grok Imagine 1.5 evidence came from Arena X, xAI docs, and an xAI employee post rather than the @xai main account.
@imagine: the Grok Imagine official account did not show a 1.5 first-release post inside the window; latest visible content remained older Extend / Agent Mode posts.
@jefffhj: the xAI employee post said Grok Imagine 1.5 Preview is available through the API, and replies said it is API only for now and will roll out to app soon. This is higher-credibility than a normal rumor, but xAI docs and actual API responses should remain the source of truth.
@GoogleDeepMind / @GeminiApp / @GoogleAI: no new model release was visible inside this window; Gemini Omni / Gemini Drops remain carryover signals from prior days.
@OpenAI: no new model release was visible inside this window; OpenAI items found in recent search were mainly Codex plan/activity/academy content, not model first releases.
@AnthropicAI / @ClaudeAI: no new model release was visible inside this window; latest model-related content remains Claude Opus 4.8 and Dynamic Workflows diffusion from prior days.
@Alibaba_Qwen / @Alibaba_Wan: no new model release was visible inside this window; Qwen3.7-Max, Wan2.7, and WanMuse+ remain prior-day signals.
@Cohere: no new model release was visible inside this window; Command A+ remains carryover from prior days.
@RunwayML: no new model release was visible inside this window; Seedance 2.0 API distribution remains carryover from prior days.
@LumaLabsAI / @Kling_ai: no new model release was visible inside this window.
@AIatMeta / @MistralAI: no new model release was visible inside this window; Meta's Muse Spark and Mistral's AI Now Summit / Vibe remain older signals.
@AlibabaCloud: the page still resolved to an anomalous title in this pass and was not used as a valid signal; future checks should continue prioritizing Qwen / Model Studio / Wan official accounts and official docs.

To Confirm

grok-imagine-video-1.5-preview: xAI docs and Arena #1 are visible, but an actual API smoke test is still needed to confirm request shape, duration/aspect/resolution limits, whether failed jobs are billed, moderation strictness, concurrency limits, and app/web rollout timing.
grok-imagine-video-1.5-preview-720p Arena score: current 1473+/-9 is still Preliminary and has only 5,564 votes; continue watching whether it holds the lead as vote count grows and whether Artificial Analysis adds a 1.5 entry.
HiDream-O1-Image-Dev-2604: Hugging Face has an MIT model card, but the page says no Inference Provider is available; local VRAM, speed, prompt-refiner effect, long-text rendering, and commercial compliance boundaries need testing.
GPT-Realtime 2.0: only the XiaoHu/Farza X demo was visible, with no OpenAI official source; wait for OpenAI Realtime API, Codex changelog, or model docs confirmation of the name and availability.
happyhorse-1.0: still strong on Arena / Artificial Analysis video leaderboards, but official model card, API, pricing, and available regions remain unclear.
Gemini Omni: user-side availability has been confirmed by multiple sources, but Gemini API / AI Studio / Vertex / Enterprise developer entry points remain unclear.
Claude Opus 4.8: Artificial Analysis LLM/GDPval already show it leading, but LMArena Text, Arena Code, and ITBench-AA still do not have complete official scores; continue watching Battle Mode conversion into leaderboard entries.
Nano Banana 2 / Pro: Google Cloud GA is confirmed, but Artificial Analysis still shows Nano Banana 2 Preview naming; continue watching whether it is replaced by GA entries and whether scores update.