Loading page
About this tool
Rank JSON, Protocol Buffers, Avro, MessagePack and CBOR against your own requirements and see why each wins or loses.
The Serialization Format Chooser ranks JSON, Protocol Buffers, Apache Avro, MessagePack and CBOR against your own weighted requirements using a decision matrix, where every fitness score is grounded in the format's specification — RFC 8259 for JSON, the protobuf language guide, the Avro 1.11 spec, the msgpack spec and RFC 8949 for CBOR. It is built for backend and platform engineers choosing a wire or storage format who want the trade-offs made explicit instead of picking whatever the last project used.
Open Serialization Format Chooser on AltFTool — it loads instantly in your browser.
Paste your code or data sample into the workspace.
Pick the format, conversion, or analysis you need.
Copy the polished result straight back into your project.
Each of the 45 fitness scores cites a documented property, like protobuf's numbered fields or CBOR's RFC 8949 constrained-node design.
Mark each requirement not needed, nice to have or critical and the ranking recomputes instantly.
Every format's ranking shows where it is weak for your requirements, not just where it wins.
Avro typically produces the smallest payloads of the mainstream formats because it writes no per-field tags at all — the reader reconstructs the record purely from the schema. Protocol Buffers is close behind but spends a varint tag on every field; JSON is usually the largest because it repeats key names as text in every record.
Use CBOR when you need an IETF standards-track format or interoperate with IoT and security stacks: CBOR (RFC 8949) is the basis of COSE signing (RFC 9052), CoAP payloads and WebAuthn attestation. MessagePack and CBOR are very similar schemaless binary encodings of the JSON data model, so outside those ecosystems the deciding factor is usually which has the better library on your platform.
No — JSON (RFC 8259) has no byte-string type, so raw binary must be base64-encoded, which adds roughly 33% to its size. Protocol Buffers, Avro, MessagePack and CBOR all have first-class binary types, which is why they score much higher when payloads contain raw bytes.
Every protobuf field has a permanent number, and parsers skip unknown numbers, so old readers accept messages with new fields and new readers accept old messages using defaults. Avro achieves comparable safety differently, by resolving the writer's schema against the reader's schema at read time; schemaless formats like JSON, MessagePack and CBOR offer no contract at all, so compatibility becomes a discipline problem.