The home of the vibe coding movement. Founded by @matthewmillerai. 111k+ on YouTube. Building to $1M in public. https://xitter.cf/t.co/wRtHrQMXKA
Joined March 2024
- Tweets9.3K
- Following20
- Followers66.9K
- Likes9.6K
The next few weeks are going to be crazy.
Fable 5.5. GPT 6.1 Astra. Grok 4.8.
And China has been quiet for a while. That's about to change.
More competition is finally coming for Opus 5.5.
BridgeMind retweeted
DeepSeek V4.1 Flash just dropped, and it's now the MOST EXPENSIVE small model on the BridgeBench lava lamp test.
DeepSeek V4.1 Flash: $0.04, 4m 3s
Claude Haiku 5.5: $0.03, 4m 34s
GPT 6 Luna: under $0.01, 42.8s
Luna built the best lamp of the three. Cheapest, fastest and best.
Also, DeepSeek's launch chart compares against Opus 5 and GPT 5.6 Sol. Two generations old.
GPT 6 Luna is still the best small model.
🚀 DeepSeek-V4.1-Flash-1017 is here!
Stronger benchmark performance, faster task completion, and shorter outputs than the previous version—a Pareto improvement in performance and cost. Now live on our website, app, and API.
🛠️ Optimized for @DeepSeekHarness, Tencent CodeBuddy, WorkBuddy, and more. Upgrade to DeepSeek Harness v0.2.1-rc.1 or later for the best experience with the new model.
1/4
Fable 5.5 needs to be affordable.
Fable 5.1 was incredible, but it died in 30 minutes on a $200 Max plan. I haven't touched it since Opus 5.5 came out.
Opus 5.5 is amazing because you can actually use it all day on a subscription.
The smartest model in the world doesn't matter if you can't afford to use it.
BridgeMind retweeted
This week's NerfBench results are live. No nerfs detected.
Claude Opus 5.5: 105.8%
Claude Sonnet 5.5: 98.9%
GPT 6.1 Sol: 98.0%
GPT 6 Astra: 97.1%
The big story is Opus 5.5. It jumped 10 points in a week, from 95.7% to 105.8%.
Last week it was the only model below its launch score. This week it's the only one above it.
Opus 5.5 is getting stronger.
I wish Opus 5.5 had more competition.
If someone actually challenged it, Anthropic would be forced to drop Fable 5.5 sooner.
But right now Opus 5.5 is still king.
So there's not much reason for them to release it.
Today we're adding frontend tests to NerfBench.
First test: Claude Opus 5.5 on the lava lamp, launch day vs today.
Sep 22: 9m 35s, 61,953 tokens.
Oct 15: 12m 39s, 78,293 tokens.
Opus 5.5 is using 26% more tokens.
But the result is way better.
Fable 5.5 is going to be a significant leap.
Muse Wonder and Gemini 4 Argon looked great on benchmarks. Neither one changed the frontier.
Fable 5.5 will.
BridgeMind retweeted
Gemini 4 Argon is now on BridgeBench.
Google put it up against GPT 6 Astra, Claude Fable 5.1, and Claude Opus 5.5 in their launch benchmarks.
On BridgeBench, it lost to all three. It lands at #8 with a 623.
It's behind Claude Sonnet 5.5, GPT 6 Sol, and even Claude Fable 5 from June.
Google's new flagship is a mid tier model.
I just cancelled my $200/month Google AI Ultra subscription.
I bought it two weeks ago to get first access to Gemini 4 Argon.
It's been a complete disappointment.
It's a step in the right direction for Google, but it doesn't compete with Opus 5.5 at all.
No reason to keep paying for it.
Gemini 4 Argon is a solid improvement for Google.
But it comes nowhere near Claude Opus 5.5.
Disappointed that Google is still this far behind.
BridgeMind retweeted
Meta's new Muse Wonder is now on BridgeBench.
It barely beats Claude Haiku 5.5. Anthropic's smallest model.
Muse Wonder: 611
Claude Haiku 5.5: 601
It's only 5 points better than Meta's own Muse Spark 1.3, and 149 points behind Claude Opus 5.5.
Meta is still nowhere near the frontier.
Muse Wonder is benchmaxxed.
Meta says it beats the competition on the key benchmarks. On the BridgeBench lava lamp test it built a red cylinder.
No glass. No lava.
I've never seen a model this benchmaxxed in my life.
BridgeMind retweeted
Latest NerfBench results are in. No nerfs detected.
Claude Opus 5.5: 95.7%
GPT 6 Astra: 102.0%
Claude Sonnet 5.5: 103.1%
GPT 6.1 Sol: 105.5%
All four are within normal variance.
But Opus 5.5 is the one to watch. It's the only model below its launch score.
We'll keep testing.
Day 238 – Vibe Coding an App Until I Make $1,000,000 | ARR: $250,772 xitter.cf/i/broadcasts/1NGaroBdq…
Claude Haiku 5.5 just dropped and it completely fails the BridgeBench lava lamp test.
No glass. No lava. Just a yellow cylinder.
GPT 6 Luna built an actual lava lamp in 42.8 seconds for under a penny.
Haiku took 4m 34s and cost $0.03.
GPT 6 Luna is over 6x faster and 3x cheaper.