Free-offer comparison
Free AI Evaluation Tools
Look at the unit behind each offer: test runs, traces, projects, or seats. Retention and team access can limit a free evaluation workflow.
Guide updated Sep 19, 2026 · Offer review and source dates appear below · Methodology
Free-offer ranking
Each row describes the provider’s recorded free offer, which may cover a different product from the category shown here. The table keeps published units separate. “Not confirmed” means we lack evidence for that detail; it does not mean the provider requires a card.
Scroll horizontally to compare every column.
| Rank | Provider | Free-access score | Recorded free offer | Access pattern | Billing/card | Offer note | Full review | Official source |
|---|---|---|---|---|---|---|---|---|
| #1 | Braintrust | 100/100 | Gateway model credit: $10 credits per month; Processed data: 1 GB per month; Scores: 10,000 scores per month | Recurring allowance | No card confirmed | Token, processed-data, and evaluation allowances have separate limits. | Jul 17, 2026 | View pricingAccessed Sep 13, 2026 |
| #2 | Future AGI | 100/100 | AI credits: 2,000 credits per month; Gateway requests: 100,000 requests per month; Voice simulation: 60 minutes per month | Recurring allowance | No card confirmed | Free usage pauses at plan limits; pay-as-you-go overages require a card. | Jul 17, 2026 | View pricingAccessed Sep 13, 2026 |
| #3 | HoneyHive | 100/100 | 10K events/mo | Recurring allowance | No card confirmed | Metrics as well as trace spans consume the monthly event allowance. | Jul 17, 2026 | View pricingAccessed Jul 17, 2026 |
| #4 | Langfuse | 100/100 | 50K units/mo | Recurring allowance | No card confirmed | A billable unit can be a trace, observation, or score; underlying model charges are separate. | Jul 17, 2026 | View pricingAccessed Sep 13, 2026 |
| #5 | Agent Jig | 95/100 | 50 agent evaluations/mo | Recurring allowance | No card confirmed | Free plan includes 50 agent evaluations per month with no credit card. | Jul 17, 2026 | View pricingAccessed Jul 17, 2026 |
| #6 | Parea AI | 90/100 | 3K logs/mo | Recurring allowance | No card confirmed | Free plan includes all platform features, two team members, 3,000 logs/month with one-month retention, and ten deployed prompts; no card required. | Jul 17, 2026 | Visit provider websiteAccessed Jul 17, 2026 |
| #7 | Arize AX / Phoenix | 75/100 | Hosted AX spans: 25,000 spans per month; Hosted AX ingestion: 1 GB per month | Recurring allowance | Not confirmed | Self-hosted Phoenix requires separate infrastructure; the hosted AX limits do not apply to it. | Jul 17, 2026 | View pricingAccessed Sep 13, 2026 |
| #8 | Weights & Biases Weave | 75/100 | Weave ingestion: 1 GB per month; Weave storage: 5 GB per month | Recurring allowance | Not confirmed | Weave ingestion and storage are separate from paid inference usage. | Jul 17, 2026 | View pricingAccessed Sep 13, 2026 |
| #9 | Galileo | 75/100 | 5K traces/mo | Recurring allowance | Not confirmed | External model or judge inference may still incur provider charges. | Jul 17, 2026 | View pricingAccessed Jul 17, 2026 |
| #10 | Helicone | 75/100 | 10K requests/mo | Recurring allowance | Not confirmed | Underlying model charges are separate. | Jul 17, 2026 | View pricingAccessed Jul 17, 2026 |
| #11 | LangSmith | 75/100 | 5K traces/mo | Recurring allowance | Not confirmed | External model calls outside included Fleet usage may be billed by the model provider. | Jul 17, 2026 | View pricingAccessed Sep 13, 2026 |
| #12 | Maxim AI | 75/100 | 10K logs/mo | Recurring allowance | Not confirmed | Development evaluations are included; production online evaluation requires a paid plan. | Jul 17, 2026 | View pricingAccessed Sep 13, 2026 |
| #13 | Traceloop | 75/100 | 50K spans/mo | Recurring allowance | Not confirmed | The hosted Free plan retains data for only 24 hours. | Jul 17, 2026 | View pricingAccessed Sep 13, 2026 |
| #14 | Bluejay | 60/100 | $25 credit | One-time balance | Not confirmed | The $25 starter credit is a one-time balance. | Jul 17, 2026 | View pricingAccessed Jul 17, 2026 |
| #15 | Roark | 55/100 | Free tier | Introductory offer | No card confirmed | The $50 starter credit is one-time; provider pass-through charges may apply. | Jul 17, 2026 | View pricingAccessed Jul 17, 2026 |
| #16 | Confident AI / DeepEval | 30/100 | Free tier | Introductory offer | Not confirmed | The hosted Free plan allows five test runs per week; additional runs are locked rather than billed as overages. | Jul 17, 2026 | View pricingAccessed Jul 17, 2026 |
Why these providers appear in this order
We score each provider’s recorded free offer: up to 40 points for recurring or durable access, 20 for a confirmed no-card signup, 15 for a stated allowance, 15 for distinct evidence links, and 10 for documented terms. Missing details earn no points. Ties favor longer-lasting access, then more detailed records, then provider name. The score does not measure product quality or whether the free offer covers this category. Read the full method before treating it as a recommendation.
Which providers are included?
We include active directory records tagged for AI evaluation tools. A tag describes a provider’s capability; its recorded free offer may cover only part of that provider’s catalog. Check the model or endpoint in the provider’s terms before building on the allowance.
See the full AI evaluation tools category, including records excluded from this guide.