Contents
Looking for a Featherless alternative? It depends which Featherless plan you’d be replacing. Featherless’s flat rate is its $25 Chat plan, and that plan is for chat you type yourself, not for apps or coding agents. For API work, Featherless bills per token, and so does SolCard Inference. The real difference is shape: Featherless gives you a catalog of 40,000+ open models; SolCard Inference gives you three open models with their refusals removed, where nothing you write is kept.
Facts about Featherless below come from its own pages, as of November 1, 2026.
At a glance
| Featherless | SolCard Inference | |
|---|---|---|
| Models | 40,000+ open models, on every plan | Three open models, refusal training removed |
| Billing for chat | Chat plan: $25 a month, flat, unlimited tokens | Pay as you go, or $20 / $100 / $200 a month |
| Billing for the API | Developer plan: per token, from $50 a month in credits | Pay as you go, or a plan once credits run out |
| Context | 32K on Chat; up to 256K on Developer | ~1M on Fast and Large; 262k on Base |
| Coding agents | Developer plan only | Every plan, and pay as you go |
| APIs | OpenAI-compatible | OpenAI Chat Completions and Responses; Anthropic Messages |
| Prompts stored | “Never” | Never |
| Pay with | Not listed on the pricing page | Crypto on 9+ networks, or a SolCard card balance |
Pricing: flat for chat, per token for everything else
Featherless’s Chat plan is $25 a month with “Unlimited tokens”, “32K context” and “4 concurrent units”: in its own words, “One 70B conversation, or four small ones at once.” The catch is in the footnote: the plan “is for interactive, human-driven use by the purchaser. Not for reselling, app/API traffic, background automation, or benchmarking.” Its table marks coding agents and production API “No” on Chat.
Its Developer plan is “Billed per token”, listed “from $50/mo”, with context “up to 256K” and 100 concurrent units. “Requests draw your balance down at each model’s own rate,” credits “Never” expire, and only successful requests are charged.
SolCard Inference bills per token from the start. On Fast, per million tokens: $1.25 input, $0.125 cached input, $2.50 output. Top up any amount and your balance never expires. Plans (Starter $20, Pro $100, Max $200 a month) buy 20% to 25% more usage than the same money as you go, and they back your API keys once your credits run out.
Neither side is unlimited for agents. Featherless caps you by concurrent units; SolCard Inference plans reset every 5 hours and every week, plan usage doesn’t roll over, and rate limits run 120 to 600 requests a minute by plan. Those are different units, so compare them on your own workload. How billing works and our pricing docs have the details.
Models: a catalog, or three picked ones
Catalog breadth is where Featherless wins, by a long way. “All 40,000+ open models on every plan,” with no model size limit, and private fine-tunes on the Developer plan. Its catalog includes community uncensored and abliterated models too. If you want to try a specific open model by name, or swap models every week, Featherless is built for that.
SolCard Inference has three models, all open-weight with their refusal training removed: Fast (the default, ~1M context), Large (most capable, always reasons, 1M context) and Base (general purpose, 262k). You don’t pick from thousands; you get three that are tuned, priced and kept running together. They’re for the work mainstream assistants refuse: security research, malware analysis for defense, red-teaming and synthetic training data. Fast, Large or Base compares them, and the models page lists their limits.
Privacy: what each one says, word for word
Featherless says it doesn’t store prompts. Its pricing page marks “Prompts stored” as “Never” on every plan, and its privacy policy says: “We do not collect or store prompts or completions.”
The same policy lists what it does keep: usage information “(i.e. # prompts, tokens in and tokens out)”, “We reserve the right to store sampler settings”, and, if you open an account, “email address and payment details.” It shares personal information with “certain service providers”, naming analytics such as Posthog. The policy doesn’t mention model training, and it was “Last updated: Monday June 10th, 2024.”
SolCard Inference makes the same core promise: prompts and responses are never logged or stored, on every model. We keep a billing line per request (time, key, model, tokens, cost) and your account’s email or wallet address. Your chat history stays in your browser unless you turn on end-to-end encrypted sync, which our servers can’t read. Keeps nothing traces one message all the way through, and seven questions shows what to ask any provider, Featherless included.
Paying
Featherless’s pricing page bills monthly and says “Cancel anytime”; it doesn’t list payment methods.
SolCard Inference takes crypto on 9+ networks (SOL, USDC, ETH, BTC and more) or a SolCard card balance, into one prepaid USD balance. There’s no card on file. Sign up with an email address or a Solana wallet.
Coding agents
Featherless’s Developer plan “Works with Cline, Kilo Code, OpenCode, OpenHands, n8n” through its OpenAI-compatible endpoint. The Chat plan isn’t for agents.
SolCard Inference speaks OpenAI Chat Completions, OpenAI Responses and Anthropic Messages, so Claude Code works natively. Paste Add SolCard Inference to my models: solcardinference.ai/install into Claude Code or Codex and our models sit next to your own. The guide and the coding agents docs cover the setup.
Which one
- Use Featherless for the widest open-model catalog, for flat-rate chat you type yourself, or to run a specific community model or fine-tune.
- Use SolCard Inference for a few models that don’t refuse, long context, native Anthropic Messages for Claude Code, and a prepaid balance you top up in crypto.
FAQ
Is Featherless pay per token?
On the Developer plan, yes: “Billed per token”, drawn from monthly credits that don’t expire. The $25 Chat plan is flat rate, but it’s limited to interactive chat by the purchaser and to 32K context.
Does Featherless log prompts?
Its privacy policy says “We do not collect or store prompts or completions.” It does record usage counts and may store sampler settings.
Can I use SolCard Inference for coding agents without a plan?
Yes. Pay as you go works with every key and every model. A plan adds more usage for the same money and backs your keys once your credits run out.