Open market for AI tokens. Compare supply, prices, and routes across providers. One API to buy and use 100+ models. discord.gg/w526yvZvFT
San Francisco
- Tweets 1,210
- Following 84
- Followers 291
- Likes 1,139
Pinned Tweet
DeepSeek is 15% cheaper on OpenModels.
> Unified API
> Just add credits and pay for the tokens you use.
🔗 openmodels.market
1
314
🚨 FLUX 3 Max takes ~12 seconds to generate a 10-second video.
The gap between quality and speed is narrowing fast.
Introducing FLUX 3 Max, the latest post-trained video model from fal Research in partnership with @bfl_ai.
FLUX 3 Max is tied for #1 in image-to-video quality on independent third-party benchmarks, generating a 10-second, 720/768p video in ~12 seconds.
For the next three weeks, FLUX 3 Max is 75% off, making it the most affordable frontier video model at its quality level.
37
Anthropic's next model. Gemini 4 Carbon. OpenAI Terra.
Which one drops first? 👀
Expect this month to be a busy one for new model releases. We have the scoop on a new Anthropic model, Google’s Gemini 4 Carbon, and a potential new Terra model from OpenAI.
theinformation.com/newslette…
43
IntentBench coming soon.
Anthropic is preparing a new AI model that can better infer what users are trying to accomplish, even when their instructions are vague.
The unreleased model also shows stronger reasoning and more creative problem-solving abilities, according to a person who has used it.
Full story: thein.fo/4irv4XQ
16
Space Bunny was MiniMax M3.1 Flash all along.
MiniMax just revealed the model after its anonymous run.
It lands at 42 on the Artificial Analysis Intelligence Index, with 1M context and pricing starting at $0.15/M input tokens.
pretty funny way to launch a model tbh.
40
🚨 BREAKING: DeepSeek V4.1 Flash 1017 just landed at #12 on WebDev Arena with 1670.
That makes it the #1 open model, ahead of Kimi K3.
Muse Wonder also enters at #16 with 1650.
DeepSeek doing this at $0.30 / $1.20 per 1M tokens is the part that really stands out.
1
1
79
How to protect your API keys with OpenModels
1. Create your keys on dashboard
2. Setup budget for those keys
3. One key routes all models
Setup budget for your every keys
26
OpenCode breach exposes a bigger problem with AI infrastructure: API keys are essentially money with an HTTP header.
Here's what happened:
> Attackers accessed a legacy system being decommissioned. Third-party API keys stored in OpenCode Console workspaces may have been exposed.
> The weak point: Legacy infrastructure with potential access to live API credentials.
> Once stolen, a bearer key can authorize requests to OpenAI, Anthropic, or Google, with the costs charged to its owner.
> OpenCode links the breach to a large-scale distillation attack affecting 1.4% of Console users.
- But there's an important distinction:
Distillation may explain what the attackers wanted. It doesn't explain how they got in.
The initial attack vector remains undisclosed.
We had a security breach last night related to a large scale distillation attack that impacted 1.4% of OpenCode Console users
We have notified them all this morning via email. If you did not get an email you were not impacted, it was a subset of users who had external providers linked
The incident is ongoing so we cannot share more details at this point but we are working with Anthropic, OpenAI, and Google to remediate
1
2
70
DeepSeek V4.1-Flash-1017 is seriously impressive.
The new update beats GPT-5.6 Sol and Claude Opus 5 on several coding and agent benchmarks, at a fraction of their API prices.
Key benchmark results:
• DeepSWE: 77.0 vs 73.0 (Sol) and 74.0 (Opus 5).
• Automation-Bench: 60.5 vs 45.8 (Sol) and 50.3 (Opus 5).
• Agents' Last Exam: 35.4 vs 26.7 (Sol) and 28.6 (Opus 5).
• Apex Bench: 72.1 vs 54.3 (Sol) and 65.8 (Opus 5).
API pricing per 1M tokens (input/output):
• DeepSeek Flash: $0.15 / $0.60 off-peak, $0.30 / $1.20 peak.
• GPT-5.6 Sol: $4 / $20.
• Claude Opus 5: $5 / $25.
These are DeepSeek's reported benchmarks, and Opus 5 still leads in several other tests.
DeepSeek also reports shorter outputs and faster task completion.
Open weights under MIT. 1M-token context window.
The price-performance gap is getting interesting.
🚨 DeepSeek V4.1 Flash 1017 just dropped.
DeepSWE jumps to 77.0, Automation-Bench to 60.5, and Apex Bench to 72.1.
DeepSeek says it also finishes tasks faster and uses shorter outputs than the previous V4.1 Flash.
quiet update, big gains.
2
305
🚨 DeepSeek V4.1 Flash 1017 just dropped.
DeepSWE jumps to 77.0, Automation-Bench to 60.5, and Apex Bench to 72.1.
DeepSeek says it also finishes tasks faster and uses shorter outputs than the previous V4.1 Flash.
quiet update, big gains.
1
191
GPT-6.1 Sol just landed at #2 on the Hermes Index.
57.09 score at just $1.19/task.
That puts it above GPT-6 Astra (56.25) while costing way less per task.
this might be the best score-to-cost combo on the board rn.
1
2
99
🚨 BREAKING: Gemini 4 Argon is now rolling out in Antigravity.
Paid Google AI Pro users are getting access today, with a separate weekly Argon quota.
ok be honest, who’s burning their Argon quota first? 👀
We’re rolling out Gemini 4 Argon in Antigravity to paid Google AI Pro subscribers today. Demand has been incredible and we've scaled capacity deliberately to get here. We’ve rolled out a separate weekly-only Argon quota for Pro on top of the existing Gemini bucket for 3.8 Flash so you have the best of Gemini for all your tasks.
64
🚨 Muse Wonder just landed on CursorBench 4.0.
> Max debuts at #12 with 50.4%
> Extra High lands at #15 with 48.2%
interesting part: Max only gains 2.2 pts over Extra High, so the extra reasoning budget is helping, but not by much here.
27
🚨 Muse Spark Open just landed on the Artificial Analysis Intelligence Index.
It debuts at #19 with a score of 39, tied with DeepSeek V4.1 Flash and just ahead of Mistral Large 4 Preview.
for an open-weight 42B active model, that’s a pretty serious first score.
Today we’re releasing Muse Spark Open, the most capable US open-weight model to date.
Built for coding and agentic work, it excels at long-horizon tasks, complex software engineering, and multi-step reasoning.
Key Model Specs:
→ 816B total / 42B active parameters
→ 1M token context window
→ Natively multimodal across text, image, video, audio
53
🚨 ChatGPT just got a big Chrome upgrade.
It can use your open tabs to compare pages, pull info, fill forms, and get things done in Chrome.
finally making tab chaos useful
91
Codex just got a fix for the "selected model is at capacity" issue.
OpenAI says it added more compute and automatic recovery, so blocked jobs should resume on their own instead of just dying there.
good. now context compaction pls 😭
Day 11/
We've landed a series of changes and more compute to very significantly decrease the occurrence of the disruptive "Selected model is at capacity. Please try a different model". When it does happen, it should also eventually just continue by itself and not just block on it.
Let us know if you still see it out there in the wild and we'll take a look.
39
Gemini 4 Argon (Autonomous 40M-line monolith refactor & bug cascade simulation)
This is ridiculously satisfying to watch
54
Gemini 4 Argon landed on CursorBench 4.0.
> High lands at #6 with 52.9%
> Medium at #7 with 52.8%
> Low at #11 with 50.6%
> Minimal at #13 with 48.8%
Google is showing up strong on coding agents already.
44