Compare supply nodes, or consume and supply idle member-session compute.
Worth examining
Supplier-set prices, site promotions, a public USD pricing endpoint, and public online Pelican tests and historical results. The sharing mechanism lets users participate in compute supply with usage-based settlement. Pelican records show a particular test; they do not establish model identity or general reliability.
Check before choosing
Multiple sources require screening. SESSION depends on both Agents and an online supplier. Public CODEX_SESSION and CLAUDE_SESSION queries around 2026-10-20 08:15 UTC both returned zero listings; confirm current supply. Aggregated price components may belong to different suppliers; budget using one selected node's full quote. The platform temporarily retains recent requests and responses; suppliers may read member-session code. Public terms list default concurrency of one and no invoices. The snapshot also reports seven-day HTTP-level group monitoring; see its scope before treating it as evidence.
Express provider choices and data policies in request settings.
Worth examining
Select or exclude providers, configure fallbacks, and set price, latency or throughput preferences. Data-collection and ZDR filters help express data requirements.
Check before choosing
Credit-purchase fees and BYOK rules affect the budget. Restricting providers or data policies narrows fallback choices. Latency and throughput preferences depend on available supply.
Access text, images, video and audio through a shared integration.
Worth examining
Multimodal API documentation, native Claude and Gemini integration guides, model mapping, fallbacks and an application ecosystem.
Check before choosing
Capabilities vary by model and protocol. Anthropic compatibility is marked Beta and should be checked separately from Claude native calls. The app-ecosystem discount excludes Claude.
Multiple protocols with usage and incident-cost records.
Worth examining
OpenAI, Anthropic and Vertex integrations, detailed usage, and compensation records organized by model and incident reason.
Check before choosing
PAYG and Builder have different permitted uses. Builder is for personal development and learning, with weekly quotas and rate limits. Full compensation thresholds and coverage conditions need further confirmation.
Recurring coding workloads and team quota management.
Worth examining
Four-week subscriptions, weekly quota refreshes, balance continuation, and team-member, key and cost management.
Check before choosing
Calculate subscription value from actual use in each week. Groups have different prices and tool restrictions; some permit Claude Code only. Convert platform dollar credit using the selected group's prices.
Use DeepSeek directly under the vendor's billing rules.
Worth examining
Official model capabilities and billing documentation provide a DeepSeek comparison reference, including cache, peak/off-peak pricing and alias rules.
Check before choosing
The model scope is concentrated on DeepSeek. Claude, GPT or Gemini need another integration. Versions, alias mappings and prices can change; retain the date.
Multimodal, embedding and reranking access, with invoice and dedicated-deployment options.
Worth examining
OpenAI/Anthropic chat protocols, published account tiers and per-model limits, invoice conditions, reserved instances and private-deployment options.
Check before choosing
Limits apply at account level, not per key; free models require identity verification. Only consumed amounts are invoiceable; account certification affects titles and invoice types. Dedicated-deployment isolation does not describe an ordinary shared API account. Documentation reviewed 2026-10-10; no calling or invoice tests.
Periodic GLM usage within supported coding tools.
Worth examining
Published input/cache/output deduction coefficients, off-peak model deductions, and vision, search and page-reading MCP features.
Check before choosing
Five-hour and weekly caps apply together; models and MCP share quota. Specified tools/endpoints only; no automatic balance continuation. Sharing, resale and relaying are prohibited. Subscriptions renew automatically and are non-refundable; cancel at least three days before renewal under reviewed guidance. Anthropic compatibility uses GLM models. Documentation reviewed 2026-10-10; no subscription or calling tests.
No matches. Try another keyword or select “All use cases”.
This guide is maintained by the TokenDos operator. TokenDos appears first. Every service uses the same source and correction standards; see the full guide for usage constraints.