Probably about Mars' size, if it was as dense as Earth.
It could be smaller that that, Mars itself would probably be a terra still if it had a magnetic field, which we know is possible even for small worlds (example: Mercury). But I doubt anything smaller than Mercury could support liquid surface water indefinitely, so Mars is probably near the lowest end of the terra size scale.
UPDATE: NASA has revised some of the stats of the Kepler-37 worlds, so I have revised the script and re-uploaded the file. Check the original post for the attachment.
All forum users, please read this! My SE mods and addons Phenom II X6 1090T 3.2 GHz, 16 GB DDR3 RAM, GTX 970 3584 MB VRAM
I think that's about right; something between Mars and Mercury sized is probably as small as a planet can get and still sustain liquid water on the surface via solar heating.
The lack of a magnetic field is generally not the main reason planets lose their atmospheres though. Direct contact with solar wind particles is indeed a cause of atmospheric loss (via sputtering), but it's often not as important as other processes such as thermal escape or impact erosion. Mars' warm wet climate appears to have come to an end about 3.8 billion years ago, right around the time of the late heavy bombardment, so impact erosion is believed to have played a significant role with large portions of the atmosphere being literally blown off the planet by massive impacts.
Additionally, a large amount of CO2 may have been absorbed by the terrain itself via rock weathering. Without active tectonics this CO2 can't get recycled back into the atmosphere, so over time this would have decreased the surface temperature. Eventually it got too cold for liquid water to exist on the surface, thus halting the rock weathering and leaving what little CO2 now remains. Meanwhile the water is almost entirely frozen or underground.
A lot of this is still quite speculative, of course, but we're learning a lot, especially thanks to Curiosity.
NASA intends to launch a mission in 2017 called the Transiting Exoplanet Survey Satellite (TESS), which will perform an all-sky survey looking for planetary transits. This will be the first space-based all-sky survey for exoplanet transits. It is expected to discover thousands of planets in the Sun's approximate neighborhood, and is designed especially to detect Earth-sized worlds in the habitable zone.
As many of you no doubt know by now, three new Kepler planets have been confirmed. Each is a super-Earth in the habitable zone.
The star Kepler-62 is a K2-type main sequence star, and is around 7 billion years old. The Kepler-62 system is 1200 light years away and features five planets. Of the three hot inner planets, two are larger than Earth, while one is about the size of Mars. The two habitable zone planets, the newfound Kepler-62 e and f, are 60% and 40% larger than Earth respectively.
Kepler-62 e is near the inner edge of the habitable zone of its sun. Its size is 60% larger than Earth. Given its mass and proximity to the star it could be a world somewhat similar to Venus. Kepler-62 f is well inside the habitable zone. At only 40% larger than Earth it is the most Earth-like world yet discovered in the habitable zone of another star. The Kepler-69 star is a G-type main sequence star slightly dimmer and cooler than Sol, and is 2700 light years away. The system contains two planets. The inner planet is a hot world more than twice the size of Earth. The outer planet, the newfound Kepler-69 c, is located on the inner edge of the habitable zone and is 70% larger than Earth.
After the announcements of the new planets today I quickly made them for SE. The procedural generation wasn't as kind to the planets as the scientists at the announcement were
Note: Currently the inclinations are rotated 90 degrees from the given Kepler data, as that value is relative to the sky (galactic equator, I think?), not the star which the planet is orbiting. The default SE extrasolar planets just use the kepler data but I believe this is incorrect. I am not 100% sure, but it can be easily changed.
I might make an "idealized" version, which modifies the planets to be more friendly in both conditions and appearance. In fact, that'd be a fun thing to do for many known exoplanets... hmm.
ENG
New site
United States 
Antarctica 
Italy 
Russian Federation 
Belgium