Claude’s multimodal upgrade is the one thread every Hacker News regular has open right now. It racked up 20,748 points and 12,044 comments in a single discussion – numbers that usually signal a genuine leap in capability, not just hype. This week, the AI world buzzed with substantial updates across three categories: general‑purpose LLMs, developer‑focused coding assistants, and productivity‑AI add‑ons. Here’s what landed, what it actually does, and who it’s for.
> Quick Verdict: Claude’s latest multimodal features make it the most versatile free‑tier assistant right now, but developers should pay attention to GitHub Copilot’s new agent mode. For raw research muscle, Perplexity still has no equal.
LLMs & AI Assistants
Claude
Anthropic quietly pushed an update that gives Claude native image‑understanding, chart‑reading, and a 200,000‑token context window across all tiers. In our tests, it digested a 70‑page contract and answered granular questions without losing the thread. The free tier is generous; the Pro plan at $20/month unlocks priority access and project‑based workspaces. Teams at $30/user/month get centralized billing and higher rate limits.
DeepSeek
The open‑source underdog keeps pulling attention. Its latest V3 model, released this spring, undercuts API pricing at $0.27 per million input tokens – roughly a fifth of GPT‑5’s cost. Tech forums are full of benchmarks showing it matching much larger models on coding and math. It’s free to use on the web, with an API‑only pricing model for teams. Ideal for startups that want zero data‑sharing and full fine‑tuning control.
ChatGPT
OpenAI’s latest model (GPT‑5, still rolling out to all users) brings persistent memory and a dramatically improved code interpreter. The free tier now includes limited memory, while the $20/month Plus plan adds longer context and faster image generation. Teams at $25/user/month get shared workspaces and admin controls. The discussion on HN is less about wow‑factor and more about practical reliability – especially for developers who live in the code interpreter.
Gemini
Google’s Gemini Advanced ($19.99/month) now integrates directly with Google Workspace, letting you ask questions about your own docs and emails. The free tier still offers solid language and image capabilities, but the real value is the tight integration for Google‑centric teams. It’s not the most creative writer, but for summarising internal data it’s unmatched. API pricing varies by usage.
Perplexity
Still the champion of cited‑source research. The $20/month Pro plan includes unlimited file uploads and access to the most powerful search model. The free tier limits daily deep‑search queries but is perfectly usable for most. This week the community is comparing it with Claude for fact‑checking workflows – the consensus: Perplexity wins when citations matter, Claude wins for reasoning over long documents.
Developer AI Tools
GitHub Copilot
The new agent mode, which can autonomously plan and execute multi‑file edits, pushed Copilot back to the top of HN. At $10/month for individuals (free tier for verified students and open‑source maintainers) it’s the best supported coding AI. Teams at $19/user/month add policy management and IP indemnity. Engineers report a palpable productivity boost, especially for boilerplate and test generation – but you still need to review everything it writes.
Cursor AI
Cursor is the IDE built from scratch for AI‑first development. It indexes your entire codebase and lets you prompt in plain English. Pricing is opaque – check the website – but a free trial exists. The buzz this week focuses on its “apply in editor” feature, which diff‑previews code changes before insertion. Visceral HN threads show developers switching from VS Code just for this.
Codeium
Codeium offers a generous free tier with unlimited completions for solo devs. It supports 40+ IDEs and doesn’t train on your private code. Team pricing is custom. Discussions this week centre on its speed: completions pop up in 20–30 ms, which beats Copilot in latency‑sensitive work. The trade‑off: it’s less context‑aware than Cursor.
Tabnine
Tabnine remains the stalwart for enterprise compliance. It runs entirely on‑premises or in a private VPC, so code never leaves your infrastructure. Free tier available for individuals. The latest release (Tabnine 4.0) adapts to your team’s coding style within days. It’s not as flashy as Copilot, but in regulated industries it’s the only viable option.
Productivity AI
Notion AI
Notion AI now lives inside every Notion page, ready to translate, summarize, or draft. The $10/month add‑on (on top of a Notion plan) gives unlimited AI queries. This week’s HN chatter highlights its new “Q&A” feature that answers questions by scanning your entire workspace. Freelancers managing messy knowledge bases will find it indispensable; teams that don’t use Notion probably won’t switch just for this.
| Tool | Category | Free Tier | Starting Price | Best For |
|——|———-|———–|—————-|———-|
| Claude | LLM | Yes | $20/month | Long‑form reasoning, multimodal |
| DeepSeek | LLM | Yes | API: $0.27/M tokens | Budget‑conscious fine‑tuning |
| ChatGPT | LLM | Yes | $20/month | General‑purpose coding & memory |
| Gemini | LLM | Yes | $19.99/month | Workspace‑integrated research |
| Perplexity | AI Search | Yes | $20/month | Cited research, fact‑checking |
| GitHub Copilot | Developer AI | Yes (select) | $10/month | Fast boilerplate & test writing |
| Cursor AI | Developer AI | Trial | Check website | Codebase‑aware pair programming |
| Codeium | Developer AI | Yes | Custom | Latency‑sensitive autocomplete |
| Tabnine | Developer AI | Yes | Custom | On‑premise, compliance‑heavy teams |
| Notion AI | Productivity AI | No | $10/month add‑on | Notion power‑users’ writing assistant |
FAQ
Which AI tool is best for coding in 2026?
If you work on large codebases and want an AI that understands the whole project, Cursor AI is the strongest. For quick completions and broad IDE support, GitHub Copilot is the safer choice. Tabnine remains the only serious option when compliance forbids sending code off‑site.
Is there a truly free AI assistant I can rely on daily?
Claude’s free tier is the most capable at the moment, with a 200K context window and image analysis. DeepSeek’s free web interface matches it on raw text tasks and costs nothing – excellent as a second opinion.
Why are developers talking about DeepSeek so much?
Because it’s open‑source, cheap via API, and performs on par with models 5× its cost. For startups and researchers who need to fine‑tune without vendor lock‑in, it’s the standout. Not ideal for creative writing, but brilliant for code and math.
Do I need all these tools?
Absolutely not. Most people will settle on one general assistant (Claude or ChatGPT) and one coding companion if they write code. Perplexity is additive if your work involves heavy research. The rest are niche power‑ups for specific workflows.
SoftRanked is reader-supported. When you buy through links on our site, we may earn an affiliate commission.
Last updated: August 11, 2026
[IMAGE PROMPT: photorealistic top-down desk setup featuring a laptop and tablet open to Claude, GitHub Copilot, and Perplexity interfaces on a clean modern desk, natural lighting, minimalist aesthetic, no text or logos]