# Anthropic launches Claude Science and a new model

An episode of [AI Chat: AI News on Claude, ChatGPT, Gemini, Grok, Perplexity](https://invisiblenetwork.ai/podcast/1669990482), 30 Jun 2026, 17 min.

Audio: https://content.rss.com/episodes/365073/2955591/ai-chat-podcast/2026_06_30_20_13_35_2de586dc-2eaa-4efc-b242-60aae5e614a2.mp3

Got something to say to this podcast? Send a voicemail, and you might hear yourself on the next episode! Claude Sonnet 5 hits 63% agentic coding at $2/M tokens; Google ships Nano Banana 2. Show Links • Get the top 80+ AI Models for $8.99 at AI Box: https://aibox.ai/builder • How I Grow and Scale My Business with AI: https://www.skool.com/aihustle • Get the AI Chat Daily Newsletter: https://www.aichatdaily.com/newsletter

## Transcript

[0:02] Welcome to the AI Chat Podcast. I'm your host,
[0:04] Jaden Schaefer. Every day I cover cutting -edge
[0:07] AI news and talk with the leaders behind it,
[0:09] breaking down what it means for your life and
[0:11] business. Anthropic is launching Claude Science.
[0:18] This is going to be a workbench for researchers.
[0:21] Base44 is launching their very own AI model.
[0:24] They're trying to protect their $100 million
[0:26] in annual recurring revenue for their vibe coding
[0:29] business. And X has just launched a hosted MCP
[0:33] server. So it's going to open its API to Claude,
[0:36] Cursor, Grok, Build, which if you know anything
[0:38] about the backstory of X and their API, this
[0:41] is quite a big deal. Claude Sonnet 5. It's going
[0:46] to be $2 per million input tokens. And Google
[0:49] is shipping Nano Banana 2 Lite and Gemini Omni
[0:52] Flash 2 developers. So tons coming out from basically
[0:55] every top AI model today. We're going to get
[0:58] into all of that. Hopefully, it's not going to
[1:00] be too noisy from my end. It is a hot, sunny
[1:03] day in North Carolina where I'm at right now,
[1:05] and I am currently bouncing a baby on my lap.
[1:08] My wife did not sleep a lot last night, so trying
[1:11] to let her get some rest, holding the baby. You
[1:14] probably heard me say this before on the podcast.
[1:16] But anyways, if you can hear a little gurgling
[1:18] and gooing in the background, that would be what
[1:21] it's from. If there's any repetitive AI tasks
[1:23] that you do, like writing newsletters or emails
[1:27] or generating graphics or anything you do on
[1:29] a repeated basis, I would love for you to go
[1:31] check out my startup, AIbox .ai. We have a no
[1:35] -code AI app builder where you basically just
[1:37] explain what you want it to do, whatever your
[1:40] workflow is, and it automatically links together
[1:42] different AI models and creates a workflow for
[1:45] you that you can use on a repeated basis. This
[1:47] saves me a lot of time because AI definitely
[1:49] is a huge... for being more productive. But at
[1:53] the same time, if you're just using it to do
[1:54] the same thing over and over every single day,
[1:56] you should probably just automate that. So if
[1:58] you want to give it a try, go to AI box dot AI
[2:01] slash builder. I'll leave a link in the description
[2:03] to go check out the builder platform that we
[2:05] have. We have over 80 different AI models that
[2:08] you can link together and build some really incredible
[2:10] automations to hopefully save you a ton of time.
[2:12] And it's only $8 .99 a month to get started.
[2:14] I'll leave a link in the description. Kicking
[2:16] it off with Anthropic. They've just launched
[2:18] Claude Science. So this is going to be a workbench.
[2:20] It's connecting 60 different scientific databases
[2:23] and a bunch of pre -built research tools for
[2:26] genomics, protein structures, chemistry. And
[2:29] all of this is going to run on Claude Opus 4
[2:31] .8. So sorry, Fable 5, not there yet. I'm sure
[2:34] when that comes out, we may get that rolled in
[2:36] as well. And it's going to come with $30 ,000
[2:39] in free credits. They're going to be giving this
[2:41] way to 50 different academic projects. And they're
[2:44] making this bet that basically workflow is going
[2:46] to beat raw capacity for getting these... researchers
[2:49] to be using Anthropic. So there's no new model
[2:53] or gating, but Claude Science uses the same Claude
[2:56] Opus 4 .8. It's basically available to everybody.
[2:59] And there is no biology fine -tuning or specialized
[3:01] weights. They've been testing this with a few
[3:03] different people and have had some impressive
[3:05] wins. They had UCSF Brain Tumor Center that was
[3:09] using it to compress geloma germline analysis
[3:12] workflows. We had the Allen Institute's Jerome
[3:15] Lecoq, who is building a computational review
[3:18] pipeline. line. And we have Novo Nordisk, who's
[3:20] one of their launch partners, what it's actually
[3:22] doing is that it's able to spin up these multi
[3:24] agent architectures. And they're going to let
[3:26] all of these scientific researchers create parallel
[3:29] sub assistants. And they're going to be doing
[3:31] sequence analysis, they're going to be doing
[3:33] structural prediction, they're going to be doing
[3:35] fact checking. And the reason why this is so
[3:36] powerful is because all of that can be happening
[3:38] while they're also keeping their full reproducibility.
[3:41] So the code, the environments, the message history,
[3:44] all of that, they're going to be able to reproduce.
[3:46] So there's a bunch of basically custom that scientists
[3:48] need. And Anthropic has built that in. They're
[3:51] giving away a bunch of grants, hoping to show
[3:53] people that this is the, you know, Anthropic
[3:55] is the best place if you're doing science. Okay,
[3:58] Base44 is launching their very own AI model.
[4:01] And the reason why is, by the way, I didn't realize,
[4:04] but Base44 is owned by Wix. It's an incredible
[4:07] vibe coding platform. I've been really impressed
[4:09] with it. It's not the one that I primarily use
[4:11] because I'm just using cloud code for mostly
[4:13] everything I do. But if you're getting started,
[4:15] I think anywhere between Loveable and Base44
[4:17] Base44, those are my two favorite. Anywhere is
[4:19] also a great platform. But in any case, Base44
[4:21] is $100 million in annual recurring revenue,
[4:24] and they just launched their brand new model
[4:26] called Base1. So this is their very own LLM.
[4:29] This is actually very similar to what Cursor
[4:30] is doing, by the way. But this LLM is trained
[4:32] on tens of millions of real user interactions.
[4:35] So what's interesting is because they're able
[4:37] to... you know, have tens of millions of users
[4:39] using them to create their tools, they can take
[4:42] all of those conversations. And I'm not sure
[4:44] if it's every single user, if they had people
[4:46] opt in, it's probably every single user, and
[4:48] it's just part of their terms of service. But
[4:49] they're taking everything that those people have
[4:51] been saying, and they're training their model
[4:53] to basically fine tune what people need when
[4:55] they're building websites, apps, and software.
[4:58] Now, what's interesting to me about this is they
[5:00] know exactly like when someone's building software,
[5:02] and you know, theoretically, anthropic and codex,
[5:05] like for opening, I should have this as well.
[5:06] But they know every time that someone asks it
[5:09] to build something and it does it, and then they
[5:11] give a follow -up, they're like, oh shoot, every
[5:12] time someone asks for this, you know, for X,
[5:14] Y, Z, they typically ask for these next five
[5:17] things. And that could either mean that when
[5:19] they first ask, they don't know exactly what
[5:21] they want, or there's traditionally a bug that
[5:23] follows, or there's a gotcha or something like
[5:25] that. But essentially they can fine tune this
[5:28] model that they have to know exactly what they
[5:31] want to, what people are basically, the direction
[5:34] they're going before they even go there, right?
[5:35] They're like, hey, I want like a fit. in this
[5:37] app and like oh these are probably the five features
[5:38] they want so either they'll ask them right away
[5:40] or they'll just start building stuff in or they'll
[5:42] set up the architecture in a way that is ready
[5:44] to go now there's a couple cool things that they
[5:46] can do with these fine -tuned models one of them
[5:48] being that they can make them much more efficient
[5:50] and so they're able to cut the costs for their
[5:52] users which of course is fantastic but also if
[5:55] someone's talking to their model and they're
[5:57] like oh you know what our you know our base one
[5:59] model would be a lot better than sending it to
[6:01] opus 4 .8 all of a sudden they're able to switch
[6:05] people over to the more efficient model, save
[6:07] the user money, and they're also keeping exclusive
[6:11] data, right? Because if they're not sending data
[6:13] over to OpenAI or Gemini or... all of a sudden
[6:17] they're not able to train with all of that data.
[6:19] So we're also seeing the same thing coming out
[6:23] of Cursor. So a bunch of different players that
[6:25] are interacting with code are basically intercepting
[6:28] the code and the conversations people are having
[6:30] and knowing what people want, and they're being
[6:32] able to train their own fine -tuned models with
[6:34] it. So base one, basically they're saying this
[6:37] is designed to replace Anthropix Opus for app
[6:40] generation workloads, and it's going to try to
[6:42] give them a really direct control over latency,
[6:44] cost, and inference speed. And they're not going
[6:46] to have to rely on any third -party APIs. And
[6:48] also, by the way, you know, this is a big cost
[6:50] for them because when you get a subscription
[6:52] to Base44, you pay them for tokens, but a huge
[6:54] chunk of your token bill just goes straight to
[6:56] Anthropic, right? It's like they're given a huge
[6:58] chunk, but if they can do it directly on their
[7:00] servers and if they have, you know, a cost -efficient
[7:03] way to do that, they're actually able to make
[7:04] more money. So Lovable that I've mentioned many
[7:07] times is, of course, their rival in Sweden, and
[7:10] they still use external LLMs. They have about
[7:12] $500 billion in AR. so they're about five times
[7:15] bigger. But they do have some really solid backing
[7:17] because Wix acquired them for about $80 million
[7:20] when they were only six months old with eight
[7:23] employees. Their parent company is now cutting
[7:25] about... 20 % of their workforce. So, you know,
[7:28] Wix is cutting down, but evidently Base44 is
[7:31] accelerating. And my prediction here is that
[7:33] if Base44 really pulls this off, Base44 will
[7:36] actually end up being much bigger than Wix, which
[7:38] is pretty wild considering they bought it for
[7:40] $80 million. It's already doing more than $100
[7:42] million in annual recurring revenue. So they
[7:45] were just, you know, blasted past even the acquisition
[7:47] price. This was an amazing deal for Wix. All
[7:50] right. The next thing we got to talk about is
[7:52] that X has just launched a hosted MCP service.
[7:55] server and if you know x like they when they
[7:57] when elon originally bought twitter he immediately
[8:00] shut down the api because i think open ai and
[8:03] a bunch of other people just had like a full
[8:05] firehose blast of all of the api data and they
[8:08] were using it for training elon shut that down
[8:10] realizing how valuable it was we saw similar
[8:12] moves from reddit and other players afterwards
[8:14] but now that x is kind of launching this new
[8:17] mcp they're going to let tools like claude cursor
[8:20] grok all tap into x ai's API, and specifically
[8:26] your own account. So if you have your own account
[8:28] or your own permissions, you can let the you
[8:31] can let the MCP tap into that. I'm sure there's
[8:33] a ton of like market sentiment analysis, you
[8:36] know, great things that you can do with this,
[8:37] there is an API. So big firms, you know, like
[8:39] hedge funds that are trying to, you know, measure
[8:42] people's sentiment analysis on the FIFA World
[8:44] Cup to determine how many hot dogs are going
[8:46] to be sold and blah, blah, blah, blah, and do
[8:48] all their like crazy, you know, hedge fund calculations,
[8:50] those types of things, they already have an API
[8:53] there. going to be using this. This is more for
[8:54] your average user. An MCP means that you can
[8:57] get all of XAI's data into your cloud that you're
[9:00] actively using. So for regular people, I personally
[9:02] would love to connect the MCP for when I talk
[9:05] about AI news here, I would love to get maybe
[9:07] like some sort of brief on every story I'm going
[9:09] to cover with the top tweets and the top comments
[9:12] on those tweets. So if my MCP could go and pull
[9:14] that, that'd be really useful. Traditionally
[9:16] trying to get like cloud to go and scrape X for
[9:19] good tweets is notoriously hard. I've had a hard
[9:22] time doing that. And I basically just have to
[9:24] go manually scroll X myself. It would be fantastic
[9:26] if I could save some time and just, you know,
[9:28] surface the best things. Basically, I mean, what's
[9:31] kind of cool is if you have an MCP into something
[9:32] like X, and I know I'm getting off the rails
[9:34] here, but you can imagine the same thing with
[9:36] Reddit or any other social platform. You can
[9:38] fine tune your own algorithm, which I think threads
[9:41] and Mark Zuckerberg's trying to let people kind
[9:43] of do that with threads. But it's like in threads
[9:45] anyways, it's like show more posts about this
[9:47] specific topic for the next like four weeks.
[9:49] And then anyways, it's not like a permanent thing.
[9:52] This would be cool. I would honestly almost remake
[9:54] my own X feed because I probably get, I'm sure
[9:56] I interact with it, but I feel like I get like
[9:58] way too much politics in my thread that I really
[10:00] want to interact with in any given day. So, and
[10:03] it's probably my own fault, but it would be really
[10:05] cool to just intentionally be able to use this
[10:07] MCP on X and say, look, I just want AI news.
[10:10] I just want, you know, news from these, you know,
[10:12] 500 companies that I'm trying to follow in AI
[10:14] and get more of a streamlined stream of content
[10:18] around that. That'd be cool for me. you know,
[10:22] the first big company to do an MCP. It feels
[10:24] like basically everyone has an MCP, even my own
[10:26] company, AIbox .ai. We have an MCP. So, you know,
[10:29] traditionally people would go to our site. There's
[10:31] 80 different AI models, audio, text, video, and
[10:34] people would go chat with them in our playground.
[10:35] And, you know, you pay like $9 a month and you
[10:37] get access to every model. But a lot of people
[10:41] are like, look, all of my workload is, all of
[10:43] my work is happening inside of Clod or inside
[10:45] of ChatGPT. So we created an MCP and now you
[10:47] can get access to all of our different tools.
[10:49] You can get, you know, like Clod can, have image
[10:51] generation and audio generation, which it didn't
[10:54] have before. So MCPs are very valuable and XAI
[10:58] is not the first one. GitHub, Slack, Stripe,
[11:00] Salesforce, all of them have official MCP endpoints.
[11:03] The API pricing, if you're going to do that with
[11:05] X is still going to be 0 .015 per published post
[11:09] and about... 20 cents per post with links. That
[11:12] was set earlier this year. They basically said
[11:14] they're trying to curb spam at scale. But the
[11:17] hosted server is showing some new capabilities.
[11:20] There's search, there's post retrieval, there's
[11:22] user lookup, there's trend analysis. All of that
[11:24] is cutting down a lot of integration time for
[11:27] people trying to work with X. So I'm going to
[11:29] be excited to test that out. Anthropic is shipping
[11:32] Claude Sonnet 5. It's going to be $2 per million
[11:36] input tokens, which is undercutting Opus 4 .8.
[11:39] And it is about... 63 .2 % on agentic coding.
[11:43] So it's within about six points of Opus 4 .8.
[11:46] You can only imagine, right, the next model,
[11:48] which is Fable 5. I mean, they already dropped
[11:51] it, but it's going to be significantly better.
[11:52] It's interesting because the, you know, their
[11:54] headline model got pulled back. So now they just
[11:56] have all of their like, all of their their worst
[11:59] models are, you know, they're more efficient
[12:00] models. They're less powerful models, they're
[12:03] starting to catch up with their main model, because
[12:05] they can't even, you know, update their main
[12:06] model to be its best, which is really interesting
[12:08] to me. So this model is going to basically become
[12:11] the default for all free and pro users starting
[12:13] on Tuesday. I think that is just showing that
[12:16] agent capable AI is now really important across
[12:19] basically every tier. This isn't going to be
[12:21] just a premium feature. And I do think that it's
[12:23] promotional pricing, the $2. for the inputs and
[12:27] the $10 for the outputs. That is going to be
[12:29] running through August 31st. And we see this
[12:31] from Anthropic a lot. Like I think if I go on
[12:33] my account right now, they're like, you know,
[12:35] the beginning of July, it's 2x credit usage.
[12:38] And I mean, basically at this point, I'm just
[12:39] trying to get my max out my usage every single
[12:42] week because I feel like you're getting a really
[12:44] great deal. And I don't know how long that will
[12:46] last, but that's going to increase from $2 to
[12:49] $3, which is still cheaper than Opus 4 .8 and
[12:52] GPT 5 .5 and Gemini 3 .1 Pro. So it's still cheaper
[12:55] but it's uh yeah until august 31st we got the
[12:58] two dollar price it's going to then go up to
[13:00] three dollars um and you know maybe we'll have
[13:02] new models coming out then so who knows what
[13:04] happens in the future but on knowledge work benchmarks
[13:07] sonnet 5 is slightly better than opus 4 .8 which
[13:11] is pretty crazy right like sonnet is the you
[13:14] know the the lower quality model than opus but
[13:17] because they can't upgrade opus now the worst
[13:19] version is actually better than opus 4 .8 i think
[13:23] that's just kind of showing though the cost per
[13:25] task efficiency is now really important it's
[13:28] more than more important than maybe the raw reasoning
[13:30] rank and if you want to spend a ton of money
[13:33] you can go use something like clod code and turn
[13:35] on like you know their super code mode or whatever
[13:38] It gives you, you know, waste more reasoning,
[13:41] but it uses a lot more tokens. A lot of early
[13:43] testers say that there are fewer mid -task stalls
[13:46] in some of the multi -step workflows. Zapier's
[13:49] two -part Salesforce automation completed where
[13:51] I think a bunch of previous models, they were
[13:53] trying to do that two -part Salesforce automation,
[13:56] and it was getting about halfway through and
[13:58] it abandoned it. So anyways, it seems like this
[14:00] is going to be much more efficient. Speaking
[14:02] of new models, they're not the only one. Google
[14:04] is shipping Nano Banana 2 Lite and also Gemini
[14:07] Omni Flash. to developers. It is a text -to -image
[14:10] model. It's going to generate in about four seconds,
[14:12] and it costs 0 .034 cents for a 1K image. Gemini
[14:17] OmniFlash, which is a video generation model,
[14:20] is about 10 cents per second of output, which
[14:24] honestly is not bad. These things can be insanely
[14:27] expensive. We literally had OpenAI that canceled
[14:29] Sora and is discontinuing it because it was just
[14:32] so expensive. So we had the capability to do
[14:35] AI video. It's just really expensive. So the
[14:37] fact that Gemini is getting that down to 10 cents
[14:39] per second of output, not bad. You know, if you
[14:42] were doing a one hour long video, it's going
[14:44] to be pretty expensive. But most of the videos
[14:47] that these are generating are like five seconds
[14:48] long. So that's 50 cents a video, which I know
[14:51] actually sounds kind of expensive. If it's 10
[14:53] second video, it's a dollar. If it's a, you know,
[14:55] 100 second video, it's $10. But you know, that's
[14:58] that's the cost. Okay, nano banana to light is
[15:01] going to be replacing the original nano banana,
[15:03] it's going to be the recommend is the recommended
[15:05] tier. And I think they're saying that it's like
[15:07] if you have a prompt that worked really well
[15:09] for the original nano banana, it's going to still
[15:11] work for nano banana to light, which is great.
[15:14] I hate it when the model changes and the prompts
[15:15] are kind of get off. They said there's going
[15:18] to be character consistency, they have really
[15:21] aggressive speed cost optimization that they've
[15:23] applied. So it's kind of interesting, right?
[15:25] It's like two, it's nano banana to light, like
[15:27] this is supposed to be like a super fast, super
[15:29] light version. And it's they're trying to get
[15:32] this thing to still be comparable to nano banana,
[15:34] but way faster and way cheaper. And I'm really
[15:36] excited when we do these kind of optimizations
[15:37] on the image models. They also have Omni flash,
[15:40] which is going to match VO three, one fast pricing
[15:44] while it's also adding conversational editing
[15:46] multimodal input mixing so they could do text
[15:49] image and video and it's going to do text to
[15:51] action synchronization for any sort of like graphics
[15:54] that you have on the screen which i think is
[15:57] So for videographers, this is actually really
[15:59] cool. Nano Banana 2 Lite is going to ship across
[16:02] nine of the different products that Google already
[16:04] has. So it's going to be inside of AI Studio.
[16:06] It's going to be inside of the Gemini API. It's
[16:08] going to be inside of AI Mode in Search. So when
[16:10] you're on Google just searching, it's going to
[16:12] be on the Gemini app, Notebook LM, Google Photos,
[16:14] Stitch, Google Flow, and then it's also inside
[16:16] of Google Ads, right? They got to make money
[16:19] from the ads off of all of this. I'm excited
[16:21] to see what Google comes out with. I'm excited
[16:23] for all the new models today. Guys, thank you
[16:24] so much for tuning into the podcast. make sure
[16:26] if you want to check out the mcp for ai box that
[16:29] you go check out ai box .ai mcp basically you're
[16:33] going to get all 80 of the top AI models right
[16:35] inside of Claude or Gemini or ChatGPT, whatever
[16:38] you work with the most. You can get all of the
[16:40] other models inside of there so you can actually
[16:42] call them. Really, really useful. And as always,
[16:44] if you want to get all of these news stories
[16:46] that I talk about on the podcast straight into
[16:48] your inbox every day, go to AIChatDaily .com.
[16:51] That's my website that goes along with this podcast.
[16:54] There's a big subscribe button you can hit in
[16:55] the top corner. And if you hit that, you can
[16:58] put in your email and I'll send you all of these
[16:59] stories and more in -depth articles. where it
[17:03] breaks down all of the information. You can see
[17:04] all of that as well. Thank you so much for tuning
[17:06] into the podcast, guys. I hope you have a fantastic
[17:08] day and I'll catch you in the next episode.

Shows are listed from Apple Podcasts' charts and directory, and Podcast Index where it is used. Every show and episode belongs to its publisher and plays from the publisher's own feed; this site hosts none of them.