In partnership with
Hey,
If you build with AI, pick tools, or just pay for a subscription, the ground shifted this past week and a half. Two of the three big labs reset their model lineups, and the headline is no longer "who's smartest." It's "who's cheapest for the job." That's genuinely good news for anyone buying.
Let's make the new map easy to read.
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
🧠 THE NEW MODEL MAP (the part that actually matters)
TL;DR: OpenAI split its flagship into three named tiers, and one of them (Terra) does GPT-5.5-level work at half the cost. Anthropic made a cheaper, more capable Claude the default. If you've been overpaying for a top-tier model to do everyday work, you probably don't need to anymore.
GPT-5.6 went public yesterday after a two-week government-gated preview. The big change is the naming: instead of "mini" and "nano," OpenAI now ships three durable tiers that will carry forward across future generations. Think of them as three points on one dial:
1- Sol · the flagship
For the hardest work
Long-horizon coding, deep reasoning, cybersecurity research, science. Two new tricks: max reasoning (more thinking time on hard problems) and ultra mode, which spins up parallel subagents to split a big task, the thing power users used to build by hand, now baked into the model. Tops the Terminal-Bench coding benchmark. Priced at $5/$30 per million tokens (same as GPT-5.5).
2- Terra · the everyday workhorse
The one most people should actually use
GPT-5.5-competitive performance at roughly half the cost. This is the sweet spot: if you've been reaching for the flagship out of habit, Terra likely does your daily work for meaningfully less. The quiet star of the launch.
3- Luna · the budget tier
For high-volume, cost-sensitive pipelines
Fast and cheap at $1/$6 per million tokens, opening a price lane below any prior OpenAI production model. If you run things at scale where "good enough, very cheap" wins, this is new territory.
And Claude has a new default. Anthropic shipped Claude Sonnet 5 on June 30, now standard on Free and Pro. The pitch mirrors Terra's: near-flagship (Opus) performance for agentic work, at a fraction of the price. Intro pricing $2/$10 per million tokens through Aug 31, then $3/$15. One thing to know if you build on it: a new tokenizer turns the same English text into up to ~35% more tokens, so the "same price" is effectively a small bump for non-Chinese content. Recount your token budget before assuming it's cheaper.
The takeaway across both: capability is now table stakes at every price point. The real question in the second half of 2026 is which tier does your specific job for the least money. Match the tier to the task and stop overpaying for the flagship by default.
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
AlphaSignal's 8-person team automated 18 workflows. No developer.
AlphaSignal runs the most-read technical AI newsletter with an 8-person team. One salesperson installed Viktor for prospect research. Sixty-seven days later it runs 18 workflows across sales, ops, editorial, and finance: proposal builder, deal updates, competitor monitoring, P&L analytics. No developer hired.
Two sentences on the partner + their link.
☝️ The ToolChase site has zero ads, zero pop-ups, zero "accept cookies to keep reading." We keep it that way by running two labeled sponsors in this email instead. If the partner above is useful to you, a click is the single easiest way to keep the site clean, at no cost to you. That's the whole model. 🙏
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
🗞️ ALSO WORTH KNOWING
Four more developments that change what you'd pick or trust:
The Claude export ban reversed. After nearly three weeks offline under a US Commerce Department order, Claude Fable 5 was restored globally on July 1 (Mythos cleared days earlier for critical infrastructure). The lesson for tool-buyers is durable: a tool you depend on can vanish by directive, not just outage. The teams that sailed through had a tested fallback.
Canva had a huge two weeks. Canva Grow 2.0 folds AI ad creation, publishing, and performance tracking into one workflow across Meta, TikTok, and LinkedIn. Canva design creation also went live inside ChatGPT and as a Codex plugin, generate and edit branded designs inside an AI conversation. Canva is quietly becoming a "make it and ship it" hub, not just a design tool.
The internet started charging the bots rent. Cloudflare launched granular control over AI crawlers (Search vs Agent vs Training), and from Sept 15 new domains block Agent and Training bots by default. Google's spam update now treats AI-answer manipulation as spam. The free-scraping era is closing; the publisher-AI relationship is shifting to paid licensing. If you make content, this decides whether AI answers cite you.
The money got historically large. VCs poured a record ~$205B into startups in Q2 2026, with AI taking an estimated 65-70% of it. Two labs alone are absorbing more capital in six months than the entire global VC market did annually a few years ago. Useful context when you're betting your workflow on a startup tool: capital concentration this extreme reshapes which tools survive.
Full running log at AI Tools News 2026.
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
🔧 THREE TOOLS WORTH A LOOK
Not ads, just directory tools that moved recently or pair with the news above:
1. Cursor — the AI code editor crossed $4B in annualized revenue and got acquired by SpaceX at a $60B valuation. Top-scored coding tool in our directory (4.8), and with the new GPT-5.6 and Sonnet 5 models landing, it's the front line where you'll actually feel the difference. If you code and you're still on a plain editor, this is the one the industry just validated with a ten-figure check.
2. Perplexity — shipped "Brain," agent memory that remembers what the agent did (what worked, what failed, your corrections) and improves overnight. Their numbers: +25% correctness on repeat tasks, -13% cost. Research preview on the Max tier, first-party metrics, but the direction, agents that get better with use, is where everything's heading.
3. Gamma — still the fastest path from idea to deck, and the free tier (400 credits, no card) genuinely gets you ~10 presentations. If the Canva-in-ChatGPT news has you rethinking your design stack, Gamma is the presentation-specific pick worth comparing against. Top-scored Design tool in our study (4.7).
All 696, ranked by editorial score, in the full directory.
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
🎬 NEW ON TOOLCHASE — Watch tools before you try them
We've embedded official product videos across tool pages, comparison pages, and listicles. The real demos, placed right where you're deciding, so you get the 60-second visual read before spending a free trial or a credit card. Reading tells you what a tool claims; watching tells you what it is.
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
📊 THE STATE OF AI TOOLS 2026
We analyzed all 696 tools in the directory. The finding that stuck with people most: paying more barely changes tool quality (free tools average 4.38, paid 4.34). Only 7% of tools are truly free, 53% are freemium, and a serious AI subscription now costs about the same as one streaming service. Nine visualizations, free to read and cite, and a fitting companion to this week's "match the tier to the task" theme.
→ Read the study · PDF · Infographic
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Same AI software. Wildly different results.
Every company in this dataset bought the same AI capabilities. The difference in results came down to one thing: whether someone inside CX owned it.
One beauty retailer made 202 workflow updates in 30 days — refining as policies changed and new questions came in. Companies without a named owner saw performance stall or decline.
Read the data on what separates AI deployments that work from the ones that stall, and the four questions worth asking before your next AI investment.
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
✍️ PROMPT OF THE WEEK — The Right-Tier Check
With three OpenAI tiers and a new default Claude, the money question is: are you using the right tier for each job? Paste this into any AI and find out:
You're a cost-optimization analyst for AI usage. Here's what I use AI for, and which model/plan I currently use for each: [LIST: task → model/tool you use now].
For each task, tell me:
1. Does this task actually need a flagship-tier model, or would a cheaper mid or budget tier do it just as well?
2. If I could downgrade, roughly how much would I save, and what (if anything) would I lose in quality?
3. Which of my tasks genuinely justify the top tier, and why?
Then give me a simple table: task, current tier, recommended tier, reason. No hype, just the math.
Why it works: most people pick one model and use it for everything, overpaying on easy tasks, occasionally under-powering hard ones. The new tiered lineups make "right tool for the job" a real lever. Ten minutes here can cut an AI bill meaningfully.
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Why we run sponsors instead of ads 🤝
Most tool directories monetize in ways we won't touch: pay-to-rank placements, invented review counts, banner ads slapped over the content. We chose the harder path. The site stays clean, the scores stay unbuyable, the reviews stay independent.
That trade only works if the newsletter carries it. So twice an issue we run a clearly-labeled sponsor, and only partners we'd genuinely recommend. If one's useful to you, a click costs nothing and tells us the model's working. No guilt, no dark patterns, just the honest mechanics of keeping a free directory free.
Want your tool in one of those two slots? Here's how it works →
That's the map. Match the tier to the task this week and see what it does to your bill.
What's the one model or tool change from the last two weeks that actually mattered to you? Reply and tell me, it hits my inbox, and I read every one.
Best,
Emre
Founder, ToolChase
P.S. Building an AI tool? Submit it for a free editorial review →
P.P.S. New here? Subscribe free → · Find us on X and YouTube.